summaryrefslogtreecommitdiff
path: root/src/vm_ppc.dasc
diff options
context:
space:
mode:
Diffstat (limited to 'src/vm_ppc.dasc')
-rw-r--r--src/vm_ppc.dasc4888
1 files changed, 4888 insertions, 0 deletions
diff --git a/src/vm_ppc.dasc b/src/vm_ppc.dasc
new file mode 100644
index 00000000..49f92d0a
--- /dev/null
+++ b/src/vm_ppc.dasc
@@ -0,0 +1,4888 @@
1|// Low-level VM code for PowerPC CPUs.
2|// Bytecode interpreter, fast functions and helper functions.
3|// Copyright (C) 2005-2012 Mike Pall. See Copyright Notice in luajit.h
4|
5|.arch ppc
6|.section code_op, code_sub
7|
8|.actionlist build_actionlist
9|.globals GLOB_
10|.globalnames globnames
11|.externnames extnames
12|
13|// Note: The ragged indentation of the instructions is intentional.
14|// The starting columns indicate data dependencies.
15|
16|//-----------------------------------------------------------------------
17|
18|// Fixed register assignments for the interpreter.
19|// Don't use: r1 = sp, r2 and r13 = reserved (TOC, TLS or SDATA)
20|
21|// The following must be C callee-save (but BASE is often refetched).
22|.define BASE, r14 // Base of current Lua stack frame.
23|.define KBASE, r15 // Constants of current Lua function.
24|.define PC, r16 // Next PC.
25|.define DISPATCH, r17 // Opcode dispatch table.
26|.define LREG, r18 // Register holding lua_State (also in SAVE_L).
27|.define MULTRES, r19 // Size of multi-result: (nresults+1)*8.
28|.define JGL, r31 // On-trace: global_State + 32768.
29|
30|// Constants for type-comparisons, stores and conversions. C callee-save.
31|.define TISNUM, r22
32|.define TISNIL, r23
33|.define ZERO, r24
34|.define TOBIT, f30 // 2^52 + 2^51.
35|.define TONUM, f31 // 2^52 + 2^51 + 2^31.
36|
37|// The following temporaries are not saved across C calls, except for RA.
38|.define RA, r20 // Callee-save.
39|.define RB, r10
40|.define RC, r11
41|.define RD, r12
42|.define INS, r7 // Overlaps CARG5.
43|
44|.define TMP0, r0
45|.define TMP1, r8
46|.define TMP2, r9
47|.define TMP3, r6 // Overlaps CARG4.
48|
49|// Saved temporaries.
50|.define SAVE0, r21
51|
52|// Calling conventions.
53|.define CARG1, r3
54|.define CARG2, r4
55|.define CARG3, r5
56|.define CARG4, r6 // Overlaps TMP3.
57|.define CARG5, r7 // Overlaps INS.
58|
59|.define FARG1, f1
60|.define FARG2, f2
61|
62|.define CRET1, r3
63|.define CRET2, r4
64|
65|// Stack layout while in interpreter. Must match with lj_frame.h.
66|.define SAVE_LR, 276(sp)
67|.define CFRAME_SPACE, 272 // Delta for sp.
68|// Back chain for sp: 272(sp) <-- sp entering interpreter
69|.define SAVE_FPR_, 128 // .. 128+18*8: 64 bit FPR saves.
70|.define SAVE_GPR_, 56 // .. 56+18*4: 32 bit GPR saves.
71|.define SAVE_CR, 52(sp) // 32 bit CR save.
72|.define SAVE_ERRF, 48(sp) // 32 bit C frame info.
73|.define SAVE_NRES, 44(sp)
74|.define SAVE_CFRAME, 40(sp)
75|.define SAVE_L, 36(sp)
76|.define SAVE_PC, 32(sp)
77|.define SAVE_MULTRES, 28(sp)
78|.define UNUSED1, 24(sp)
79|.define TMPD_LO, 20(sp)
80|.define TMPD_HI, 16(sp)
81|.define TONUM_LO, 12(sp)
82|.define TONUM_HI, 8(sp)
83|// Next frame lr: 4(sp)
84|// Back chain for sp: 0(sp) <-- sp while in interpreter
85|
86|.define TMPD_BLO, 23(sp)
87|.define TMPD, TMPD_HI
88|.define TONUM_D, TONUM_HI
89|
90|.macro save_, reg
91| stw r..reg, SAVE_GPR_+(reg-14)*4(sp)
92| stfd f..reg, SAVE_FPR_+(reg-14)*8(sp)
93|.endmacro
94|.macro rest_, reg
95| lwz r..reg, SAVE_GPR_+(reg-14)*4(sp)
96| lfd f..reg, SAVE_FPR_+(reg-14)*8(sp)
97|.endmacro
98|
99|.macro saveregs
100| stwu sp, -CFRAME_SPACE(sp)
101| save_ 14; save_ 15; save_ 16
102| mflr r0
103| save_ 17; save_ 18; save_ 19; save_ 20; save_ 21; save_ 22
104| stw r0, SAVE_LR
105| save_ 23; save_ 24; save_ 25
106| mfcr r0
107| save_ 26; save_ 27; save_ 28; save_ 29; save_ 30; save_ 31
108| stw r0, SAVE_CR
109|.endmacro
110|
111|.macro restoreregs
112| lwz r0, SAVE_LR; lwz r12, SAVE_CR
113| rest_ 14; rest_ 15; rest_ 16; rest_ 17; rest_ 18; rest_ 19
114| mtlr r0; mtcrf 0x38, r12
115| rest_ 20; rest_ 21; rest_ 22; rest_ 23; rest_ 24; rest_ 25
116| rest_ 26; rest_ 27; rest_ 28; rest_ 29; rest_ 30; rest_ 31
117| addi sp, sp, CFRAME_SPACE
118|.endmacro
119|
120|// Type definitions. Some of these are only used for documentation.
121|.type L, lua_State, LREG
122|.type GL, global_State
123|.type TVALUE, TValue
124|.type GCOBJ, GCobj
125|.type STR, GCstr
126|.type TAB, GCtab
127|.type LFUNC, GCfuncL
128|.type CFUNC, GCfuncC
129|.type PROTO, GCproto
130|.type UPVAL, GCupval
131|.type NODE, Node
132|.type NARGS8, int
133|.type TRACE, GCtrace
134|
135|//-----------------------------------------------------------------------
136|
137|// These basic macros should really be part of DynASM.
138|.macro srwi, rx, ry, n; rlwinm rx, ry, 32-n, n, 31; .endmacro
139|.macro slwi, rx, ry, n; rlwinm rx, ry, n, 0, 31-n; .endmacro
140|.macro rotlwi, rx, ry, n; rlwinm rx, ry, n, 0, 31; .endmacro
141|.macro rotlw, rx, ry, rn; rlwnm rx, ry, rn, 0, 31; .endmacro
142|.macro subi, rx, ry, i; addi rx, ry, -i; .endmacro
143|
144|// Trap for not-yet-implemented parts.
145|.macro NYI; tw 4, sp, sp; .endmacro
146|
147|// int/FP conversions.
148|.macro tonum_i, freg, reg
149| xoris reg, reg, 0x8000
150| stw reg, TONUM_LO
151| lfd freg, TONUM_D
152| fsub freg, freg, TONUM
153|.endmacro
154|
155|.macro tonum_u, freg, reg
156| stw reg, TONUM_LO
157| lfd freg, TONUM_D
158| fsub freg, freg, TOBIT
159|.endmacro
160|
161|.macro toint, reg, freg, tmpfreg
162| fctiwz tmpfreg, freg
163| stfd tmpfreg, TMPD
164| lwz reg, TMPD_LO
165|.endmacro
166|
167|.macro toint, reg, freg
168| toint reg, freg, freg
169|.endmacro
170|
171|//-----------------------------------------------------------------------
172|
173|// Access to frame relative to BASE.
174|.define FRAME_PC, -8
175|.define FRAME_FUNC, -4
176|
177|// Instruction decode.
178|.macro decode_OP4, dst, ins; rlwinm dst, ins, 2, 22, 29; .endmacro
179|.macro decode_RA8, dst, ins; rlwinm dst, ins, 27, 21, 28; .endmacro
180|.macro decode_RB8, dst, ins; rlwinm dst, ins, 11, 21, 28; .endmacro
181|.macro decode_RC8, dst, ins; rlwinm dst, ins, 19, 21, 28; .endmacro
182|.macro decode_RD8, dst, ins; rlwinm dst, ins, 19, 13, 28; .endmacro
183|
184|.macro decode_OP1, dst, ins; rlwinm dst, ins, 0, 24, 31; .endmacro
185|.macro decode_RD4, dst, ins; rlwinm dst, ins, 18, 14, 29; .endmacro
186|
187|// Instruction fetch.
188|.macro ins_NEXT1
189| lwz INS, 0(PC)
190| addi PC, PC, 4
191|.endmacro
192|// Instruction decode+dispatch. Note: optimized for e300!
193|.macro ins_NEXT2
194| decode_OP4 TMP1, INS
195| lwzx TMP0, DISPATCH, TMP1
196| mtctr TMP0
197| decode_RB8 RB, INS
198| decode_RD8 RD, INS
199| decode_RA8 RA, INS
200| decode_RC8 RC, INS
201| bctr
202|.endmacro
203|.macro ins_NEXT
204| ins_NEXT1
205| ins_NEXT2
206|.endmacro
207|
208|// Instruction footer.
209|.if 1
210| // Replicated dispatch. Less unpredictable branches, but higher I-Cache use.
211| .define ins_next, ins_NEXT
212| .define ins_next_, ins_NEXT
213| .define ins_next1, ins_NEXT1
214| .define ins_next2, ins_NEXT2
215|.else
216| // Common dispatch. Lower I-Cache use, only one (very) unpredictable branch.
217| // Affects only certain kinds of benchmarks (and only with -j off).
218| .macro ins_next
219| b ->ins_next
220| .endmacro
221| .macro ins_next1
222| .endmacro
223| .macro ins_next2
224| b ->ins_next
225| .endmacro
226| .macro ins_next_
227| ->ins_next:
228| ins_NEXT
229| .endmacro
230|.endif
231|
232|// Call decode and dispatch.
233|.macro ins_callt
234| // BASE = new base, RB = LFUNC/CFUNC, RC = nargs*8, FRAME_PC(BASE) = PC
235| lwz PC, LFUNC:RB->pc
236| lwz INS, 0(PC)
237| addi PC, PC, 4
238| decode_OP4 TMP1, INS
239| decode_RA8 RA, INS
240| lwzx TMP0, DISPATCH, TMP1
241| add RA, RA, BASE
242| mtctr TMP0
243| bctr
244|.endmacro
245|
246|.macro ins_call
247| // BASE = new base, RB = LFUNC/CFUNC, RC = nargs*8, PC = caller PC
248| stw PC, FRAME_PC(BASE)
249| ins_callt
250|.endmacro
251|
252|//-----------------------------------------------------------------------
253|
254|// Macros to test operand types.
255|.macro checknum, reg; cmplw reg, TISNUM; .endmacro
256|.macro checknum, cr, reg; cmplw cr, reg, TISNUM; .endmacro
257|.macro checkstr, reg; cmpwi reg, LJ_TSTR; .endmacro
258|.macro checktab, reg; cmpwi reg, LJ_TTAB; .endmacro
259|.macro checkfunc, reg; cmpwi reg, LJ_TFUNC; .endmacro
260|.macro checknil, reg; cmpwi reg, LJ_TNIL; .endmacro
261|
262|.macro branch_RD
263| srwi TMP0, RD, 1
264| addis PC, PC, -(BCBIAS_J*4 >> 16)
265| add PC, PC, TMP0
266|.endmacro
267|
268|// Assumes DISPATCH is relative to GL.
269#define DISPATCH_GL(field) (GG_DISP2G + (int)offsetof(global_State, field))
270#define DISPATCH_J(field) (GG_DISP2J + (int)offsetof(jit_State, field))
271|
272#define PC2PROTO(field) ((int)offsetof(GCproto, field)-(int)sizeof(GCproto))
273|
274|.macro hotcheck, delta, target
275| rlwinm TMP1, PC, 31, 25, 30
276| addi TMP1, TMP1, GG_DISP2HOT
277| lhzx TMP2, DISPATCH, TMP1
278| addic. TMP2, TMP2, -delta
279| sthx TMP2, DISPATCH, TMP1
280| blt target
281|.endmacro
282|
283|.macro hotloop
284| hotcheck HOTCOUNT_LOOP, ->vm_hotloop
285|.endmacro
286|
287|.macro hotcall
288| hotcheck HOTCOUNT_CALL, ->vm_hotcall
289|.endmacro
290|
291|// Set current VM state. Uses TMP0.
292|.macro li_vmstate, st; li TMP0, ~LJ_VMST_..st; .endmacro
293|.macro st_vmstate; stw TMP0, DISPATCH_GL(vmstate)(DISPATCH); .endmacro
294|
295|// Move table write barrier back. Overwrites mark and tmp.
296|.macro barrierback, tab, mark, tmp
297| lwz tmp, DISPATCH_GL(gc.grayagain)(DISPATCH)
298| // Assumes LJ_GC_BLACK is 0x04.
299| rlwinm mark, mark, 0, 30, 28 // black2gray(tab)
300| stw tab, DISPATCH_GL(gc.grayagain)(DISPATCH)
301| stb mark, tab->marked
302| stw tmp, tab->gclist
303|.endmacro
304|
305|//-----------------------------------------------------------------------
306
307/* Generate subroutines used by opcodes and other parts of the VM. */
308/* The .code_sub section should be last to help static branch prediction. */
309static void build_subroutines(BuildCtx *ctx)
310{
311 |.code_sub
312 |
313 |//-----------------------------------------------------------------------
314 |//-- Return handling ----------------------------------------------------
315 |//-----------------------------------------------------------------------
316 |
317 |->vm_returnp:
318 | // See vm_return. Also: TMP2 = previous base.
319 | andi. TMP0, PC, FRAME_P
320 | li TMP1, LJ_TTRUE
321 | beq ->cont_dispatch
322 |
323 | // Return from pcall or xpcall fast func.
324 | lwz PC, FRAME_PC(TMP2) // Fetch PC of previous frame.
325 | mr BASE, TMP2 // Restore caller base.
326 | // Prepending may overwrite the pcall frame, so do it at the end.
327 | stwu TMP1, FRAME_PC(RA) // Prepend true to results.
328 |
329 |->vm_returnc:
330 | andi. TMP0, PC, FRAME_TYPE
331 | addi RD, RD, 8 // RD = (nresults+1)*8.
332 | mr MULTRES, RD
333 | beq ->BC_RET_Z // Handle regular return to Lua.
334 |
335 |->vm_return:
336 | // BASE = base, RA = resultptr, RD/MULTRES = (nresults+1)*8, PC = return
337 | // TMP0 = PC & FRAME_TYPE
338 | cmpwi TMP0, FRAME_C
339 | rlwinm TMP2, PC, 0, 0, 28
340 | li_vmstate C
341 | sub TMP2, BASE, TMP2 // TMP2 = previous base.
342 | bney ->vm_returnp
343 |
344 | addic. TMP1, RD, -8
345 | stw TMP2, L->base
346 | lwz TMP2, SAVE_NRES
347 | subi BASE, BASE, 8
348 | st_vmstate
349 | slwi TMP2, TMP2, 3
350 | beq >2
351 |1:
352 | addic. TMP1, TMP1, -8
353 | lfd f0, 0(RA)
354 | addi RA, RA, 8
355 | stfd f0, 0(BASE)
356 | addi BASE, BASE, 8
357 | bney <1
358 |
359 |2:
360 | cmpw TMP2, RD // More/less results wanted?
361 | bne >6
362 |3:
363 | stw BASE, L->top // Store new top.
364 |
365 |->vm_leave_cp:
366 | lwz TMP0, SAVE_CFRAME // Restore previous C frame.
367 | li CRET1, 0 // Ok return status for vm_pcall.
368 | stw TMP0, L->cframe
369 |
370 |->vm_leave_unw:
371 | restoreregs
372 | blr
373 |
374 |6:
375 | ble >7 // Less results wanted?
376 | // More results wanted. Check stack size and fill up results with nil.
377 | lwz TMP1, L->maxstack
378 | cmplw BASE, TMP1
379 | bge >8
380 | stw TISNIL, 0(BASE)
381 | addi RD, RD, 8
382 | addi BASE, BASE, 8
383 | b <2
384 |
385 |7: // Less results wanted.
386 | subfic TMP3, TMP2, 0 // LUA_MULTRET+1 case?
387 | sub TMP0, RD, TMP2
388 | subfe TMP1, TMP1, TMP1 // TMP1 = TMP2 == 0 ? 0 : -1
389 | and TMP0, TMP0, TMP1
390 | sub BASE, BASE, TMP0 // Either keep top or shrink it.
391 | b <3
392 |
393 |8: // Corner case: need to grow stack for filling up results.
394 | // This can happen if:
395 | // - A C function grows the stack (a lot).
396 | // - The GC shrinks the stack in between.
397 | // - A return back from a lua_call() with (high) nresults adjustment.
398 | stw BASE, L->top // Save current top held in BASE (yes).
399 | mr SAVE0, RD
400 | mr CARG2, TMP2
401 | mr CARG1, L
402 | bl extern lj_state_growstack // (lua_State *L, int n)
403 | lwz TMP2, SAVE_NRES
404 | mr RD, SAVE0
405 | slwi TMP2, TMP2, 3
406 | lwz BASE, L->top // Need the (realloced) L->top in BASE.
407 | b <2
408 |
409 |->vm_unwind_c: // Unwind C stack, return from vm_pcall.
410 | // (void *cframe, int errcode)
411 | mr sp, CARG1
412 | mr CRET1, CARG2
413 |->vm_unwind_c_eh: // Landing pad for external unwinder.
414 | lwz L, SAVE_L
415 | li TMP0, ~LJ_VMST_C
416 | lwz GL:TMP1, L->glref
417 | stw TMP0, GL:TMP1->vmstate
418 | b ->vm_leave_unw
419 |
420 |->vm_unwind_ff: // Unwind C stack, return from ff pcall.
421 | // (void *cframe)
422 | rlwinm sp, CARG1, 0, 0, 29
423 |->vm_unwind_ff_eh: // Landing pad for external unwinder.
424 | lwz L, SAVE_L
425 | li TISNUM, LJ_TISNUM // Setup type comparison constants.
426 | lwz BASE, L->base
427 | lus TMP3, 0x59c0 // TOBIT = 2^52 + 2^51 (float).
428 | lwz DISPATCH, L->glref // Setup pointer to dispatch table.
429 | li ZERO, 0
430 | stw TMP3, TMPD
431 | li TMP1, LJ_TFALSE
432 | ori TMP3, TMP3, 0x0004 // TONUM = 2^52 + 2^51 + 2^31 (float).
433 | li TISNIL, LJ_TNIL
434 | li_vmstate INTERP
435 | lfs TOBIT, TMPD
436 | lwz PC, FRAME_PC(BASE) // Fetch PC of previous frame.
437 | la RA, -8(BASE) // Results start at BASE-8.
438 | stw TMP3, TMPD
439 | addi DISPATCH, DISPATCH, GG_G2DISP
440 | stw TMP1, 0(RA) // Prepend false to error message.
441 | li RD, 16 // 2 results: false + error message.
442 | st_vmstate
443 | lfs TONUM, TMPD
444 | b ->vm_returnc
445 |
446 |//-----------------------------------------------------------------------
447 |//-- Grow stack for calls -----------------------------------------------
448 |//-----------------------------------------------------------------------
449 |
450 |->vm_growstack_c: // Grow stack for C function.
451 | li CARG2, LUA_MINSTACK
452 | b >2
453 |
454 |->vm_growstack_l: // Grow stack for Lua function.
455 | // BASE = new base, RA = BASE+framesize*8, RC = nargs*8, PC = first PC
456 | add RC, BASE, RC
457 | sub RA, RA, BASE
458 | stw BASE, L->base
459 | addi PC, PC, 4 // Must point after first instruction.
460 | stw RC, L->top
461 | srwi CARG2, RA, 3
462 |2:
463 | // L->base = new base, L->top = top
464 | stw PC, SAVE_PC
465 | mr CARG1, L
466 | bl extern lj_state_growstack // (lua_State *L, int n)
467 | lwz BASE, L->base
468 | lwz RC, L->top
469 | lwz LFUNC:RB, FRAME_FUNC(BASE)
470 | sub RC, RC, BASE
471 | // BASE = new base, RB = LFUNC/CFUNC, RC = nargs*8, FRAME_PC(BASE) = PC
472 | ins_callt // Just retry the call.
473 |
474 |//-----------------------------------------------------------------------
475 |//-- Entry points into the assembler VM ---------------------------------
476 |//-----------------------------------------------------------------------
477 |
478 |->vm_resume: // Setup C frame and resume thread.
479 | // (lua_State *L, TValue *base, int nres1 = 0, ptrdiff_t ef = 0)
480 | saveregs
481 | mr L, CARG1
482 | lwz DISPATCH, L->glref // Setup pointer to dispatch table.
483 | mr BASE, CARG2
484 | lbz TMP1, L->status
485 | stw L, SAVE_L
486 | li PC, FRAME_CP
487 | addi TMP0, sp, CFRAME_RESUME
488 | addi DISPATCH, DISPATCH, GG_G2DISP
489 | stw CARG3, SAVE_NRES
490 | cmplwi TMP1, 0
491 | stw CARG3, SAVE_ERRF
492 | stw TMP0, L->cframe
493 | stw CARG3, SAVE_CFRAME
494 | stw CARG1, SAVE_PC // Any value outside of bytecode is ok.
495 | beq >3
496 |
497 | // Resume after yield (like a return).
498 | mr RA, BASE
499 | lwz BASE, L->base
500 | li TISNUM, LJ_TISNUM // Setup type comparison constants.
501 | lwz TMP1, L->top
502 | lwz PC, FRAME_PC(BASE)
503 | lus TMP3, 0x59c0 // TOBIT = 2^52 + 2^51 (float).
504 | stb CARG3, L->status
505 | stw TMP3, TMPD
506 | ori TMP3, TMP3, 0x0004 // TONUM = 2^52 + 2^51 + 2^31 (float).
507 | lfs TOBIT, TMPD
508 | sub RD, TMP1, BASE
509 | stw TMP3, TMPD
510 | lus TMP0, 0x4338 // Hiword of 2^52 + 2^51 (double)
511 | addi RD, RD, 8
512 | stw TMP0, TONUM_HI
513 | li_vmstate INTERP
514 | li ZERO, 0
515 | st_vmstate
516 | andi. TMP0, PC, FRAME_TYPE
517 | mr MULTRES, RD
518 | lfs TONUM, TMPD
519 | li TISNIL, LJ_TNIL
520 | beq ->BC_RET_Z
521 | b ->vm_return
522 |
523 |->vm_pcall: // Setup protected C frame and enter VM.
524 | // (lua_State *L, TValue *base, int nres1, ptrdiff_t ef)
525 | saveregs
526 | li PC, FRAME_CP
527 | stw CARG4, SAVE_ERRF
528 | b >1
529 |
530 |->vm_call: // Setup C frame and enter VM.
531 | // (lua_State *L, TValue *base, int nres1)
532 | saveregs
533 | li PC, FRAME_C
534 |
535 |1: // Entry point for vm_pcall above (PC = ftype).
536 | lwz TMP1, L:CARG1->cframe
537 | stw CARG3, SAVE_NRES
538 | mr L, CARG1
539 | stw CARG1, SAVE_L
540 | mr BASE, CARG2
541 | stw sp, L->cframe // Add our C frame to cframe chain.
542 | lwz DISPATCH, L->glref // Setup pointer to dispatch table.
543 | stw CARG1, SAVE_PC // Any value outside of bytecode is ok.
544 | stw TMP1, SAVE_CFRAME
545 | addi DISPATCH, DISPATCH, GG_G2DISP
546 |
547 |3: // Entry point for vm_cpcall/vm_resume (BASE = base, PC = ftype).
548 | lwz TMP2, L->base // TMP2 = old base (used in vmeta_call).
549 | li TISNUM, LJ_TISNUM // Setup type comparison constants.
550 | lwz TMP1, L->top
551 | lus TMP3, 0x59c0 // TOBIT = 2^52 + 2^51 (float).
552 | add PC, PC, BASE
553 | stw TMP3, TMPD
554 | li ZERO, 0
555 | ori TMP3, TMP3, 0x0004 // TONUM = 2^52 + 2^51 + 2^31 (float).
556 | lfs TOBIT, TMPD
557 | sub PC, PC, TMP2 // PC = frame delta + frame type
558 | stw TMP3, TMPD
559 | lus TMP0, 0x4338 // Hiword of 2^52 + 2^51 (double)
560 | sub NARGS8:RC, TMP1, BASE
561 | stw TMP0, TONUM_HI
562 | li_vmstate INTERP
563 | lfs TONUM, TMPD
564 | li TISNIL, LJ_TNIL
565 | st_vmstate
566 |
567 |->vm_call_dispatch:
568 | // TMP2 = old base, BASE = new base, RC = nargs*8, PC = caller PC
569 | lwz TMP0, FRAME_PC(BASE)
570 | lwz LFUNC:RB, FRAME_FUNC(BASE)
571 | checkfunc TMP0; bne ->vmeta_call
572 |
573 |->vm_call_dispatch_f:
574 | ins_call
575 | // BASE = new base, RB = func, RC = nargs*8, PC = caller PC
576 |
577 |->vm_cpcall: // Setup protected C frame, call C.
578 | // (lua_State *L, lua_CFunction func, void *ud, lua_CPFunction cp)
579 | saveregs
580 | mr L, CARG1
581 | lwz TMP0, L:CARG1->stack
582 | stw CARG1, SAVE_L
583 | lwz TMP1, L->top
584 | stw CARG1, SAVE_PC // Any value outside of bytecode is ok.
585 | sub TMP0, TMP0, TMP1 // Compute -savestack(L, L->top).
586 | lwz TMP1, L->cframe
587 | stw sp, L->cframe // Add our C frame to cframe chain.
588 | li TMP2, 0
589 | stw TMP0, SAVE_NRES // Neg. delta means cframe w/o frame.
590 | stw TMP2, SAVE_ERRF // No error function.
591 | stw TMP1, SAVE_CFRAME
592 | mtctr CARG4
593 | bctrl // (lua_State *L, lua_CFunction func, void *ud)
594 | mr. BASE, CRET1
595 | lwz DISPATCH, L->glref // Setup pointer to dispatch table.
596 | li PC, FRAME_CP
597 | addi DISPATCH, DISPATCH, GG_G2DISP
598 | bne <3 // Else continue with the call.
599 | b ->vm_leave_cp // No base? Just remove C frame.
600 |
601 |//-----------------------------------------------------------------------
602 |//-- Metamethod handling ------------------------------------------------
603 |//-----------------------------------------------------------------------
604 |
605 |// The lj_meta_* functions (except for lj_meta_cat) don't reallocate the
606 |// stack, so BASE doesn't need to be reloaded across these calls.
607 |
608 |//-- Continuation dispatch ----------------------------------------------
609 |
610 |->cont_dispatch:
611 | // BASE = meta base, RA = resultptr, RD = (nresults+1)*8
612 | lwz TMP0, -12(BASE) // Continuation.
613 | mr RB, BASE
614 | mr BASE, TMP2 // Restore caller BASE.
615 | lwz LFUNC:TMP1, FRAME_FUNC(TMP2)
616#if LJ_HASFFI
617 | cmplwi TMP0, 1
618#endif
619 | lwz PC, -16(RB) // Restore PC from [cont|PC].
620 | subi TMP2, RD, 8
621 | lwz TMP1, LFUNC:TMP1->pc
622 | stwx TISNIL, RA, TMP2 // Ensure one valid arg.
623#if LJ_HASFFI
624 | ble >1
625#endif
626 | lwz KBASE, PC2PROTO(k)(TMP1)
627 | // BASE = base, RA = resultptr, RB = meta base
628 | mtctr TMP0
629 | bctr // Jump to continuation.
630 |
631#if LJ_HASFFI
632 |1:
633 | beq ->cont_ffi_callback // cont = 1: return from FFI callback.
634 | // cont = 0: tailcall from C function.
635 | subi TMP1, RB, 16
636 | sub RC, TMP1, BASE
637 | b ->vm_call_tail
638#endif
639 |
640 |->cont_cat: // RA = resultptr, RB = meta base
641 | lwz INS, -4(PC)
642 | subi CARG2, RB, 16
643 | decode_RB8 SAVE0, INS
644 | lfd f0, 0(RA)
645 | add TMP1, BASE, SAVE0
646 | stw BASE, L->base
647 | cmplw TMP1, CARG2
648 | sub CARG3, CARG2, TMP1
649 | decode_RA8 RA, INS
650 | stfd f0, 0(CARG2)
651 | bney ->BC_CAT_Z
652 | stfdx f0, BASE, RA
653 | b ->cont_nop
654 |
655 |//-- Table indexing metamethods -----------------------------------------
656 |
657 |->vmeta_tgets1:
658 | la CARG3, DISPATCH_GL(tmptv)(DISPATCH)
659 | li TMP0, LJ_TSTR
660 | decode_RB8 RB, INS
661 | stw STR:RC, 4(CARG3)
662 | add CARG2, BASE, RB
663 | stw TMP0, 0(CARG3)
664 | b >1
665 |
666 |->vmeta_tgets:
667 | la CARG2, DISPATCH_GL(tmptv)(DISPATCH)
668 | li TMP0, LJ_TTAB
669 | stw TAB:RB, 4(CARG2)
670 | la CARG3, DISPATCH_GL(tmptv2)(DISPATCH)
671 | stw TMP0, 0(CARG2)
672 | li TMP1, LJ_TSTR
673 | stw STR:RC, 4(CARG3)
674 | stw TMP1, 0(CARG3)
675 | b >1
676 |
677 |->vmeta_tgetb: // TMP0 = index
678 if (!LJ_DUALNUM) {
679 | tonum_u f0, TMP0
680 }
681 | decode_RB8 RB, INS
682 | la CARG3, DISPATCH_GL(tmptv)(DISPATCH)
683 | add CARG2, BASE, RB
684 if (LJ_DUALNUM) {
685 | stw TISNUM, 0(CARG3)
686 | stw TMP0, 4(CARG3)
687 } else {
688 | stfd f0, 0(CARG3)
689 }
690 | b >1
691 |
692 |->vmeta_tgetv:
693 | decode_RB8 RB, INS
694 | decode_RC8 RC, INS
695 | add CARG2, BASE, RB
696 | add CARG3, BASE, RC
697 |1:
698 | stw BASE, L->base
699 | mr CARG1, L
700 | stw PC, SAVE_PC
701 | bl extern lj_meta_tget // (lua_State *L, TValue *o, TValue *k)
702 | // Returns TValue * (finished) or NULL (metamethod).
703 | cmplwi CRET1, 0
704 | beq >3
705 | lfd f0, 0(CRET1)
706 | ins_next1
707 | stfdx f0, BASE, RA
708 | ins_next2
709 |
710 |3: // Call __index metamethod.
711 | // BASE = base, L->top = new base, stack = cont/func/t/k
712 | subfic TMP1, BASE, FRAME_CONT
713 | lwz BASE, L->top
714 | stw PC, -16(BASE) // [cont|PC]
715 | add PC, TMP1, BASE
716 | lwz LFUNC:RB, FRAME_FUNC(BASE) // Guaranteed to be a function here.
717 | li NARGS8:RC, 16 // 2 args for func(t, k).
718 | b ->vm_call_dispatch_f
719 |
720 |//-----------------------------------------------------------------------
721 |
722 |->vmeta_tsets1:
723 | la CARG3, DISPATCH_GL(tmptv)(DISPATCH)
724 | li TMP0, LJ_TSTR
725 | decode_RB8 RB, INS
726 | stw STR:RC, 4(CARG3)
727 | add CARG2, BASE, RB
728 | stw TMP0, 0(CARG3)
729 | b >1
730 |
731 |->vmeta_tsets:
732 | la CARG2, DISPATCH_GL(tmptv)(DISPATCH)
733 | li TMP0, LJ_TTAB
734 | stw TAB:RB, 4(CARG2)
735 | la CARG3, DISPATCH_GL(tmptv2)(DISPATCH)
736 | stw TMP0, 0(CARG2)
737 | li TMP1, LJ_TSTR
738 | stw STR:RC, 4(CARG3)
739 | stw TMP1, 0(CARG3)
740 | b >1
741 |
742 |->vmeta_tsetb: // TMP0 = index
743 if (!LJ_DUALNUM) {
744 | tonum_u f0, TMP0
745 }
746 | decode_RB8 RB, INS
747 | la CARG3, DISPATCH_GL(tmptv)(DISPATCH)
748 | add CARG2, BASE, RB
749 if (LJ_DUALNUM) {
750 | stw TISNUM, 0(CARG3)
751 | stw TMP0, 4(CARG3)
752 } else {
753 | stfd f0, 0(CARG3)
754 }
755 | b >1
756 |
757 |->vmeta_tsetv:
758 | decode_RB8 RB, INS
759 | decode_RC8 RC, INS
760 | add CARG2, BASE, RB
761 | add CARG3, BASE, RC
762 |1:
763 | stw BASE, L->base
764 | mr CARG1, L
765 | stw PC, SAVE_PC
766 | bl extern lj_meta_tset // (lua_State *L, TValue *o, TValue *k)
767 | // Returns TValue * (finished) or NULL (metamethod).
768 | cmplwi CRET1, 0
769 | lfdx f0, BASE, RA
770 | beq >3
771 | // NOBARRIER: lj_meta_tset ensures the table is not black.
772 | ins_next1
773 | stfd f0, 0(CRET1)
774 | ins_next2
775 |
776 |3: // Call __newindex metamethod.
777 | // BASE = base, L->top = new base, stack = cont/func/t/k/(v)
778 | subfic TMP1, BASE, FRAME_CONT
779 | lwz BASE, L->top
780 | stw PC, -16(BASE) // [cont|PC]
781 | add PC, TMP1, BASE
782 | lwz LFUNC:RB, FRAME_FUNC(BASE) // Guaranteed to be a function here.
783 | li NARGS8:RC, 24 // 3 args for func(t, k, v)
784 | stfd f0, 16(BASE) // Copy value to third argument.
785 | b ->vm_call_dispatch_f
786 |
787 |//-- Comparison metamethods ---------------------------------------------
788 |
789 |->vmeta_comp:
790 | mr CARG1, L
791 | subi PC, PC, 4
792 if (LJ_DUALNUM) {
793 | mr CARG2, RA
794 } else {
795 | add CARG2, BASE, RA
796 }
797 | stw PC, SAVE_PC
798 if (LJ_DUALNUM) {
799 | mr CARG3, RD
800 } else {
801 | add CARG3, BASE, RD
802 }
803 | stw BASE, L->base
804 | decode_OP1 CARG4, INS
805 | bl extern lj_meta_comp // (lua_State *L, TValue *o1, *o2, int op)
806 | // Returns 0/1 or TValue * (metamethod).
807 |3:
808 | cmplwi CRET1, 1
809 | bgt ->vmeta_binop
810 | subfic CRET1, CRET1, 0
811 |4:
812 | lwz INS, 0(PC)
813 | addi PC, PC, 4
814 | decode_RD4 TMP2, INS
815 | addis TMP2, TMP2, -(BCBIAS_J*4 >> 16)
816 | and TMP2, TMP2, CRET1
817 | add PC, PC, TMP2
818 |->cont_nop:
819 | ins_next
820 |
821 |->cont_ra: // RA = resultptr
822 | lwz INS, -4(PC)
823 | lfd f0, 0(RA)
824 | decode_RA8 TMP1, INS
825 | stfdx f0, BASE, TMP1
826 | b ->cont_nop
827 |
828 |->cont_condt: // RA = resultptr
829 | lwz TMP0, 0(RA)
830 | subfic TMP0, TMP0, LJ_TTRUE // Branch if result is true.
831 | subfe CRET1, CRET1, CRET1
832 | not CRET1, CRET1
833 | b <4
834 |
835 |->cont_condf: // RA = resultptr
836 | lwz TMP0, 0(RA)
837 | subfic TMP0, TMP0, LJ_TTRUE // Branch if result is false.
838 | subfe CRET1, CRET1, CRET1
839 | b <4
840 |
841 |->vmeta_equal:
842 | // CARG2, CARG3, CARG4 are already set by BC_ISEQV/BC_ISNEV.
843 | subi PC, PC, 4
844 | stw BASE, L->base
845 | mr CARG1, L
846 | stw PC, SAVE_PC
847 | bl extern lj_meta_equal // (lua_State *L, GCobj *o1, *o2, int ne)
848 | // Returns 0/1 or TValue * (metamethod).
849 | b <3
850 |
851 |->vmeta_equal_cd:
852#if LJ_HASFFI
853 | mr CARG2, INS
854 | subi PC, PC, 4
855 | stw BASE, L->base
856 | mr CARG1, L
857 | stw PC, SAVE_PC
858 | bl extern lj_meta_equal_cd // (lua_State *L, BCIns op)
859 | // Returns 0/1 or TValue * (metamethod).
860 | b <3
861#endif
862 |
863 |//-- Arithmetic metamethods ---------------------------------------------
864 |
865 |->vmeta_arith_nv:
866 | add CARG3, KBASE, RC
867 | add CARG4, BASE, RB
868 | b >1
869 |->vmeta_arith_nv2:
870 if (LJ_DUALNUM) {
871 | mr CARG3, RC
872 | mr CARG4, RB
873 | b >1
874 }
875 |
876 |->vmeta_unm:
877 | mr CARG3, RD
878 | mr CARG4, RD
879 | b >1
880 |
881 |->vmeta_arith_vn:
882 | add CARG3, BASE, RB
883 | add CARG4, KBASE, RC
884 | b >1
885 |
886 |->vmeta_arith_vv:
887 | add CARG3, BASE, RB
888 | add CARG4, BASE, RC
889 if (LJ_DUALNUM) {
890 | b >1
891 }
892 |->vmeta_arith_vn2:
893 |->vmeta_arith_vv2:
894 if (LJ_DUALNUM) {
895 | mr CARG3, RB
896 | mr CARG4, RC
897 }
898 |1:
899 | add CARG2, BASE, RA
900 | stw BASE, L->base
901 | mr CARG1, L
902 | stw PC, SAVE_PC
903 | decode_OP1 CARG5, INS // Caveat: CARG5 overlaps INS.
904 | bl extern lj_meta_arith // (lua_State *L, TValue *ra,*rb,*rc, BCReg op)
905 | // Returns NULL (finished) or TValue * (metamethod).
906 | cmplwi CRET1, 0
907 | beq ->cont_nop
908 |
909 | // Call metamethod for binary op.
910 |->vmeta_binop:
911 | // BASE = old base, CRET1 = new base, stack = cont/func/o1/o2
912 | sub TMP1, CRET1, BASE
913 | stw PC, -16(CRET1) // [cont|PC]
914 | mr TMP2, BASE
915 | addi PC, TMP1, FRAME_CONT
916 | mr BASE, CRET1
917 | li NARGS8:RC, 16 // 2 args for func(o1, o2).
918 | b ->vm_call_dispatch
919 |
920 |->vmeta_len:
921#ifdef LUAJIT_ENABLE_LUA52COMPAT
922 | mr SAVE0, CARG1
923#endif
924 | mr CARG2, RD
925 | stw BASE, L->base
926 | mr CARG1, L
927 | stw PC, SAVE_PC
928 | bl extern lj_meta_len // (lua_State *L, TValue *o)
929 | // Returns NULL (retry) or TValue * (metamethod base).
930#ifdef LUAJIT_ENABLE_LUA52COMPAT
931 | cmplwi CRET1, 0
932 | bne ->vmeta_binop // Binop call for compatibility.
933 | mr CARG1, SAVE0
934 | b ->BC_LEN_Z
935#else
936 | b ->vmeta_binop // Binop call for compatibility.
937#endif
938 |
939 |//-- Call metamethod ----------------------------------------------------
940 |
941 |->vmeta_call: // Resolve and call __call metamethod.
942 | // TMP2 = old base, BASE = new base, RC = nargs*8
943 | mr CARG1, L
944 | stw TMP2, L->base // This is the callers base!
945 | subi CARG2, BASE, 8
946 | stw PC, SAVE_PC
947 | add CARG3, BASE, RC
948 | mr SAVE0, NARGS8:RC
949 | bl extern lj_meta_call // (lua_State *L, TValue *func, TValue *top)
950 | lwz LFUNC:RB, FRAME_FUNC(BASE) // Guaranteed to be a function here.
951 | addi NARGS8:RC, SAVE0, 8 // Got one more argument now.
952 | ins_call
953 |
954 |->vmeta_callt: // Resolve __call for BC_CALLT.
955 | // BASE = old base, RA = new base, RC = nargs*8
956 | mr CARG1, L
957 | stw BASE, L->base
958 | subi CARG2, RA, 8
959 | stw PC, SAVE_PC
960 | add CARG3, RA, RC
961 | mr SAVE0, NARGS8:RC
962 | bl extern lj_meta_call // (lua_State *L, TValue *func, TValue *top)
963 | lwz TMP1, FRAME_PC(BASE)
964 | addi NARGS8:RC, SAVE0, 8 // Got one more argument now.
965 | lwz LFUNC:RB, FRAME_FUNC(RA) // Guaranteed to be a function here.
966 | b ->BC_CALLT_Z
967 |
968 |//-- Argument coercion for 'for' statement ------------------------------
969 |
970 |->vmeta_for:
971 | mr CARG1, L
972 | stw BASE, L->base
973 | mr CARG2, RA
974 | stw PC, SAVE_PC
975 | mr SAVE0, INS
976 | bl extern lj_meta_for // (lua_State *L, TValue *base)
977#if LJ_HASJIT
978 | decode_OP1 TMP0, SAVE0
979#endif
980 | decode_RA8 RA, SAVE0
981#if LJ_HASJIT
982 | cmpwi TMP0, BC_JFORI
983#endif
984 | decode_RD8 RD, SAVE0
985#if LJ_HASJIT
986 | beqy =>BC_JFORI
987#endif
988 | b =>BC_FORI
989 |
990 |//-----------------------------------------------------------------------
991 |//-- Fast functions -----------------------------------------------------
992 |//-----------------------------------------------------------------------
993 |
994 |.macro .ffunc, name
995 |->ff_ .. name:
996 |.endmacro
997 |
998 |.macro .ffunc_1, name
999 |->ff_ .. name:
1000 | cmplwi NARGS8:RC, 8
1001 | lwz CARG3, 0(BASE)
1002 | lwz CARG1, 4(BASE)
1003 | blt ->fff_fallback
1004 |.endmacro
1005 |
1006 |.macro .ffunc_2, name
1007 |->ff_ .. name:
1008 | cmplwi NARGS8:RC, 16
1009 | lwz CARG3, 0(BASE)
1010 | lwz CARG4, 8(BASE)
1011 | lwz CARG1, 4(BASE)
1012 | lwz CARG2, 12(BASE)
1013 | blt ->fff_fallback
1014 |.endmacro
1015 |
1016 |.macro .ffunc_n, name
1017 |->ff_ .. name:
1018 | cmplwi NARGS8:RC, 8
1019 | lwz CARG3, 0(BASE)
1020 | lfd FARG1, 0(BASE)
1021 | blt ->fff_fallback
1022 | checknum CARG3; bge ->fff_fallback
1023 |.endmacro
1024 |
1025 |.macro .ffunc_nn, name
1026 |->ff_ .. name:
1027 | cmplwi NARGS8:RC, 16
1028 | lwz CARG3, 0(BASE)
1029 | lfd FARG1, 0(BASE)
1030 | lwz CARG4, 8(BASE)
1031 | lfd FARG2, 8(BASE)
1032 | blt ->fff_fallback
1033 | checknum CARG3; bge ->fff_fallback
1034 | checknum CARG4; bge ->fff_fallback
1035 |.endmacro
1036 |
1037 |// Inlined GC threshold check. Caveat: uses TMP0 and TMP1.
1038 |.macro ffgccheck
1039 | lwz TMP0, DISPATCH_GL(gc.total)(DISPATCH)
1040 | lwz TMP1, DISPATCH_GL(gc.threshold)(DISPATCH)
1041 | cmplw TMP0, TMP1
1042 | bgel ->fff_gcstep
1043 |.endmacro
1044 |
1045 |//-- Base library: checks -----------------------------------------------
1046 |
1047 |.ffunc_1 assert
1048 | li TMP1, LJ_TFALSE
1049 | la RA, -8(BASE)
1050 | cmplw cr1, CARG3, TMP1
1051 | lwz PC, FRAME_PC(BASE)
1052 | bge cr1, ->fff_fallback
1053 | stw CARG3, 0(RA)
1054 | addi RD, NARGS8:RC, 8 // Compute (nresults+1)*8.
1055 | stw CARG1, 4(RA)
1056 | beq ->fff_res // Done if exactly 1 argument.
1057 | li TMP1, 8
1058 | subi RC, RC, 8
1059 |1:
1060 | cmplw TMP1, RC
1061 | lfdx f0, BASE, TMP1
1062 | stfdx f0, RA, TMP1
1063 | addi TMP1, TMP1, 8
1064 | bney <1
1065 | b ->fff_res
1066 |
1067 |.ffunc type
1068 | cmplwi NARGS8:RC, 8
1069 | lwz CARG1, 0(BASE)
1070 | blt ->fff_fallback
1071 | subfc TMP0, TISNUM, CARG1
1072 | subfe TMP2, CARG1, CARG1
1073 | orc TMP1, TMP2, TMP0
1074 | addi TMP1, TMP1, ~LJ_TISNUM+1
1075 | slwi TMP1, TMP1, 3
1076 | la TMP2, CFUNC:RB->upvalue
1077 | lfdx FARG1, TMP2, TMP1
1078 | b ->fff_resn
1079 |
1080 |//-- Base library: getters and setters ---------------------------------
1081 |
1082 |.ffunc_1 getmetatable
1083 | checktab CARG3; bne >6
1084 |1: // Field metatable must be at same offset for GCtab and GCudata!
1085 | lwz TAB:CARG1, TAB:CARG1->metatable
1086 |2:
1087 | li CARG3, LJ_TNIL
1088 | cmplwi TAB:CARG1, 0
1089 | lwz STR:RC, DISPATCH_GL(gcroot[GCROOT_MMNAME+MM_metatable])(DISPATCH)
1090 | beq ->fff_restv
1091 | lwz TMP0, TAB:CARG1->hmask
1092 | li CARG3, LJ_TTAB // Use metatable as default result.
1093 | lwz TMP1, STR:RC->hash
1094 | lwz NODE:TMP2, TAB:CARG1->node
1095 | and TMP1, TMP1, TMP0 // idx = str->hash & tab->hmask
1096 | slwi TMP0, TMP1, 5
1097 | slwi TMP1, TMP1, 3
1098 | sub TMP1, TMP0, TMP1
1099 | add NODE:TMP2, NODE:TMP2, TMP1 // node = tab->node + (idx*32-idx*8)
1100 |3: // Rearranged logic, because we expect _not_ to find the key.
1101 | lwz CARG4, NODE:TMP2->key
1102 | lwz TMP0, 4+offsetof(Node, key)(NODE:TMP2)
1103 | lwz CARG2, NODE:TMP2->val
1104 | lwz TMP1, 4+offsetof(Node, val)(NODE:TMP2)
1105 | checkstr CARG4; bne >4
1106 | cmpw TMP0, STR:RC; beq >5
1107 |4:
1108 | lwz NODE:TMP2, NODE:TMP2->next
1109 | cmplwi NODE:TMP2, 0
1110 | beq ->fff_restv // Not found, keep default result.
1111 | b <3
1112 |5:
1113 | checknil CARG2
1114 | beq ->fff_restv // Ditto for nil value.
1115 | mr CARG3, CARG2 // Return value of mt.__metatable.
1116 | mr CARG1, TMP1
1117 | b ->fff_restv
1118 |
1119 |6:
1120 | cmpwi CARG3, LJ_TUDATA; beq <1
1121 | subfc TMP0, TISNUM, CARG3
1122 | subfe TMP2, CARG3, CARG3
1123 | orc TMP1, TMP2, TMP0
1124 | addi TMP1, TMP1, ~LJ_TISNUM+1
1125 | slwi TMP1, TMP1, 2
1126 | la TMP2, DISPATCH_GL(gcroot[GCROOT_BASEMT])(DISPATCH)
1127 | lwzx TAB:CARG1, TMP2, TMP1
1128 | b <2
1129 |
1130 |.ffunc_2 setmetatable
1131 | // Fast path: no mt for table yet and not clearing the mt.
1132 | checktab CARG3; bne ->fff_fallback
1133 | lwz TAB:TMP1, TAB:CARG1->metatable
1134 | checktab CARG4; bne ->fff_fallback
1135 | cmplwi TAB:TMP1, 0
1136 | lbz TMP3, TAB:CARG1->marked
1137 | bne ->fff_fallback
1138 | andi. TMP0, TMP3, LJ_GC_BLACK // isblack(table)
1139 | stw TAB:CARG2, TAB:CARG1->metatable
1140 | beq ->fff_restv
1141 | barrierback TAB:CARG1, TMP3, TMP0
1142 | b ->fff_restv
1143 |
1144 |.ffunc rawget
1145 | cmplwi NARGS8:RC, 16
1146 | lwz CARG4, 0(BASE)
1147 | lwz TAB:CARG2, 4(BASE)
1148 | blt ->fff_fallback
1149 | checktab CARG4; bne ->fff_fallback
1150 | la CARG3, 8(BASE)
1151 | mr CARG1, L
1152 | bl extern lj_tab_get // (lua_State *L, GCtab *t, cTValue *key)
1153 | // Returns cTValue *.
1154 | lfd FARG1, 0(CRET1)
1155 | b ->fff_resn
1156 |
1157 |//-- Base library: conversions ------------------------------------------
1158 |
1159 |.ffunc tonumber
1160 | // Only handles the number case inline (without a base argument).
1161 | cmplwi NARGS8:RC, 8
1162 | lwz CARG1, 0(BASE)
1163 | lfd FARG1, 0(BASE)
1164 | bne ->fff_fallback // Exactly one argument.
1165 | checknum CARG1; bgt ->fff_fallback
1166 | b ->fff_resn
1167 |
1168 |.ffunc_1 tostring
1169 | // Only handles the string or number case inline.
1170 | checkstr CARG3
1171 | // A __tostring method in the string base metatable is ignored.
1172 | beq ->fff_restv // String key?
1173 | // Handle numbers inline, unless a number base metatable is present.
1174 | lwz TMP0, DISPATCH_GL(gcroot[GCROOT_BASEMT_NUM])(DISPATCH)
1175 | checknum CARG3
1176 | cmplwi cr1, TMP0, 0
1177 | stw BASE, L->base // Add frame since C call can throw.
1178 | crorc 4*cr0+eq, 4*cr0+gt, 4*cr1+eq
1179 | stw PC, SAVE_PC // Redundant (but a defined value).
1180 | beq ->fff_fallback
1181 | ffgccheck
1182 | mr CARG1, L
1183 | mr CARG2, BASE
1184 if (LJ_DUALNUM) {
1185 | bl extern lj_str_fromnumber // (lua_State *L, cTValue *o)
1186 } else {
1187 | bl extern lj_str_fromnum // (lua_State *L, lua_Number *np)
1188 }
1189 | // Returns GCstr *.
1190 | li CARG3, LJ_TSTR
1191 | b ->fff_restv
1192 |
1193 |//-- Base library: iterators -------------------------------------------
1194 |
1195 |.ffunc next
1196 | cmplwi NARGS8:RC, 8
1197 | lwz CARG1, 0(BASE)
1198 | lwz TAB:CARG2, 4(BASE)
1199 | blt ->fff_fallback
1200 | stwx TISNIL, BASE, NARGS8:RC // Set missing 2nd arg to nil.
1201 | checktab CARG1
1202 | lwz PC, FRAME_PC(BASE)
1203 | bne ->fff_fallback
1204 | stw BASE, L->base // Add frame since C call can throw.
1205 | mr CARG1, L
1206 | stw BASE, L->top // Dummy frame length is ok.
1207 | la CARG3, 8(BASE)
1208 | stw PC, SAVE_PC
1209 | bl extern lj_tab_next // (lua_State *L, GCtab *t, TValue *key)
1210 | // Returns 0 at end of traversal.
1211 | cmplwi CRET1, 0
1212 | li CARG3, LJ_TNIL
1213 | beq ->fff_restv // End of traversal: return nil.
1214 | lfd f0, 8(BASE) // Copy key and value to results.
1215 | la RA, -8(BASE)
1216 | lfd f1, 16(BASE)
1217 | stfd f0, 0(RA)
1218 | li RD, (2+1)*8
1219 | stfd f1, 8(RA)
1220 | b ->fff_res
1221 |
1222 |.ffunc_1 pairs
1223 | checktab CARG3
1224 | lwz PC, FRAME_PC(BASE)
1225 | bne ->fff_fallback
1226#ifdef LUAJIT_ENABLE_LUA52COMPAT
1227 | lwz TAB:TMP2, TAB:CARG1->metatable
1228 | lfd f0, CFUNC:RB->upvalue[0]
1229 | cmplwi TAB:TMP2, 0
1230 | la RA, -8(BASE)
1231 | bne ->fff_fallback
1232#else
1233 | lfd f0, CFUNC:RB->upvalue[0]
1234 | la RA, -8(BASE)
1235#endif
1236 | stw TISNIL, 8(BASE)
1237 | li RD, (3+1)*8
1238 | stfd f0, 0(RA)
1239 | b ->fff_res
1240 |
1241 |.ffunc ipairs_aux
1242 | cmplwi NARGS8:RC, 16
1243 | lwz CARG3, 0(BASE)
1244 | lwz TAB:CARG1, 4(BASE)
1245 | lwz CARG4, 8(BASE)
1246 if (LJ_DUALNUM) {
1247 | lwz TMP2, 12(BASE)
1248 } else {
1249 | lfd FARG2, 8(BASE)
1250 }
1251 | blt ->fff_fallback
1252 | checktab CARG3
1253 | checknum cr1, CARG4
1254 | lwz PC, FRAME_PC(BASE)
1255 if (LJ_DUALNUM) {
1256 | bne ->fff_fallback
1257 | bne cr1, ->fff_fallback
1258 } else {
1259 | lus TMP0, 0x3ff0
1260 | stw ZERO, TMPD_LO
1261 | bne ->fff_fallback
1262 | stw TMP0, TMPD_HI
1263 | bge cr1, ->fff_fallback
1264 | lfd FARG1, TMPD
1265 | toint TMP2, FARG2, f0
1266 }
1267 | lwz TMP0, TAB:CARG1->asize
1268 | lwz TMP1, TAB:CARG1->array
1269 if (!LJ_DUALNUM) {
1270 | fadd FARG2, FARG2, FARG1
1271 }
1272 | addi TMP2, TMP2, 1
1273 | la RA, -8(BASE)
1274 | cmplw TMP0, TMP2
1275 if (LJ_DUALNUM) {
1276 | stw TISNUM, 0(RA)
1277 | slwi TMP3, TMP2, 3
1278 | stw TMP2, 4(RA)
1279 } else {
1280 | slwi TMP3, TMP2, 3
1281 | stfd FARG2, 0(RA)
1282 }
1283 | ble >2 // Not in array part?
1284 | lwzx TMP2, TMP1, TMP3
1285 | lfdx f0, TMP1, TMP3
1286 |1:
1287 | checknil TMP2
1288 | li RD, (0+1)*8
1289 | beq ->fff_res // End of iteration, return 0 results.
1290 | li RD, (2+1)*8
1291 | stfd f0, 8(RA)
1292 | b ->fff_res
1293 |2: // Check for empty hash part first. Otherwise call C function.
1294 | lwz TMP0, TAB:CARG1->hmask
1295 | cmplwi TMP0, 0
1296 | li RD, (0+1)*8
1297 | beq ->fff_res
1298 | mr CARG2, TMP2
1299 | bl extern lj_tab_getinth // (GCtab *t, int32_t key)
1300 | // Returns cTValue * or NULL.
1301 | cmplwi CRET1, 0
1302 | li RD, (0+1)*8
1303 | beq ->fff_res
1304 | lwz TMP2, 0(CRET1)
1305 | lfd f0, 0(CRET1)
1306 | b <1
1307 |
1308 |.ffunc_1 ipairs
1309 | checktab CARG3
1310 | lwz PC, FRAME_PC(BASE)
1311 | bne ->fff_fallback
1312#ifdef LUAJIT_ENABLE_LUA52COMPAT
1313 | lwz TAB:TMP2, TAB:CARG1->metatable
1314 | lfd f0, CFUNC:RB->upvalue[0]
1315 | cmplwi TAB:TMP2, 0
1316 | la RA, -8(BASE)
1317 | bne ->fff_fallback
1318#else
1319 | lfd f0, CFUNC:RB->upvalue[0]
1320 | la RA, -8(BASE)
1321#endif
1322 if (LJ_DUALNUM) {
1323 | stw TISNUM, 8(BASE)
1324 } else {
1325 | stw ZERO, 8(BASE)
1326 }
1327 | stw ZERO, 12(BASE)
1328 | li RD, (3+1)*8
1329 | stfd f0, 0(RA)
1330 | b ->fff_res
1331 |
1332 |//-- Base library: catch errors ----------------------------------------
1333 |
1334 |.ffunc pcall
1335 | cmplwi NARGS8:RC, 8
1336 | lbz TMP3, DISPATCH_GL(hookmask)(DISPATCH)
1337 | blt ->fff_fallback
1338 | mr TMP2, BASE
1339 | la BASE, 8(BASE)
1340 | // Remember active hook before pcall.
1341 | rlwinm TMP3, TMP3, 32-HOOK_ACTIVE_SHIFT, 31, 31
1342 | subi NARGS8:RC, NARGS8:RC, 8
1343 | addi PC, TMP3, 8+FRAME_PCALL
1344 | b ->vm_call_dispatch
1345 |
1346 |.ffunc xpcall
1347 | cmplwi NARGS8:RC, 16
1348 | lwz CARG4, 8(BASE)
1349 | lfd FARG2, 8(BASE)
1350 | lfd FARG1, 0(BASE)
1351 | blt ->fff_fallback
1352 | lbz TMP1, DISPATCH_GL(hookmask)(DISPATCH)
1353 | mr TMP2, BASE
1354 | checkfunc CARG4; bne ->fff_fallback // Traceback must be a function.
1355 | la BASE, 16(BASE)
1356 | // Remember active hook before pcall.
1357 | rlwinm TMP1, TMP1, 32-HOOK_ACTIVE_SHIFT, 31, 31
1358 | stfd FARG2, 0(TMP2) // Swap function and traceback.
1359 | subi NARGS8:RC, NARGS8:RC, 16
1360 | stfd FARG1, 8(TMP2)
1361 | addi PC, TMP1, 16+FRAME_PCALL
1362 | b ->vm_call_dispatch
1363 |
1364 |//-- Coroutine library --------------------------------------------------
1365 |
1366 |.macro coroutine_resume_wrap, resume
1367 |.if resume
1368 |.ffunc_1 coroutine_resume
1369 | cmpwi CARG3, LJ_TTHREAD; bne ->fff_fallback
1370 |.else
1371 |.ffunc coroutine_wrap_aux
1372 | lwz L:CARG1, CFUNC:RB->upvalue[0].gcr
1373 |.endif
1374 | lbz TMP0, L:CARG1->status
1375 | lwz TMP1, L:CARG1->cframe
1376 | lwz CARG2, L:CARG1->top
1377 | cmplwi cr0, TMP0, LUA_YIELD
1378 | lwz TMP2, L:CARG1->base
1379 | cmplwi cr1, TMP1, 0
1380 | lwz TMP0, L:CARG1->maxstack
1381 | cmplw cr7, CARG2, TMP2
1382 | lwz PC, FRAME_PC(BASE)
1383 | crorc 4*cr6+lt, 4*cr0+gt, 4*cr1+eq // st>LUA_YIELD || cframe!=0
1384 | add TMP2, CARG2, NARGS8:RC
1385 | crandc 4*cr6+gt, 4*cr7+eq, 4*cr0+eq // base==top && st!=LUA_YIELD
1386 | cmplw cr1, TMP2, TMP0
1387 | cror 4*cr6+lt, 4*cr6+lt, 4*cr6+gt
1388 | stw PC, SAVE_PC
1389 | cror 4*cr6+lt, 4*cr6+lt, 4*cr1+gt // cond1 || cond2 || stackov
1390 | stw BASE, L->base
1391 | blt cr6, ->fff_fallback
1392 |1:
1393 |.if resume
1394 | addi BASE, BASE, 8 // Keep resumed thread in stack for GC.
1395 | subi NARGS8:RC, NARGS8:RC, 8
1396 | subi TMP2, TMP2, 8
1397 |.endif
1398 | stw TMP2, L:CARG1->top
1399 | li TMP1, 0
1400 | stw BASE, L->top
1401 |2: // Move args to coroutine.
1402 | cmpw TMP1, NARGS8:RC
1403 | lfdx f0, BASE, TMP1
1404 | beq >3
1405 | stfdx f0, CARG2, TMP1
1406 | addi TMP1, TMP1, 8
1407 | b <2
1408 |3:
1409 | li CARG3, 0
1410 | mr L:SAVE0, L:CARG1
1411 | li CARG4, 0
1412 | bl ->vm_resume // (lua_State *L, TValue *base, 0, 0)
1413 | // Returns thread status.
1414 |4:
1415 | lwz TMP2, L:SAVE0->base
1416 | cmplwi CRET1, LUA_YIELD
1417 | lwz TMP3, L:SAVE0->top
1418 | li_vmstate INTERP
1419 | lwz BASE, L->base
1420 | st_vmstate
1421 | bgt >8
1422 | sub RD, TMP3, TMP2
1423 | lwz TMP0, L->maxstack
1424 | cmplwi RD, 0
1425 | add TMP1, BASE, RD
1426 | beq >6 // No results?
1427 | cmplw TMP1, TMP0
1428 | li TMP1, 0
1429 | bgt >9 // Need to grow stack?
1430 |
1431 | subi TMP3, RD, 8
1432 | stw TMP2, L:SAVE0->top // Clear coroutine stack.
1433 |5: // Move results from coroutine.
1434 | cmplw TMP1, TMP3
1435 | lfdx f0, TMP2, TMP1
1436 | stfdx f0, BASE, TMP1
1437 | addi TMP1, TMP1, 8
1438 | bne <5
1439 |6:
1440 | andi. TMP0, PC, FRAME_TYPE
1441 |.if resume
1442 | li TMP1, LJ_TTRUE
1443 | la RA, -8(BASE)
1444 | stw TMP1, -8(BASE) // Prepend true to results.
1445 | addi RD, RD, 16
1446 |.else
1447 | mr RA, BASE
1448 | addi RD, RD, 8
1449 |.endif
1450 |7:
1451 | stw PC, SAVE_PC
1452 | mr MULTRES, RD
1453 | beq ->BC_RET_Z
1454 | b ->vm_return
1455 |
1456 |8: // Coroutine returned with error (at co->top-1).
1457 |.if resume
1458 | andi. TMP0, PC, FRAME_TYPE
1459 | la TMP3, -8(TMP3)
1460 | li TMP1, LJ_TFALSE
1461 | lfd f0, 0(TMP3)
1462 | stw TMP3, L:SAVE0->top // Remove error from coroutine stack.
1463 | li RD, (2+1)*8
1464 | stw TMP1, -8(BASE) // Prepend false to results.
1465 | la RA, -8(BASE)
1466 | stfd f0, 0(BASE) // Copy error message.
1467 | b <7
1468 |.else
1469 | mr CARG1, L
1470 | mr CARG2, L:SAVE0
1471 | bl extern lj_ffh_coroutine_wrap_err // (lua_State *L, lua_State *co)
1472 |.endif
1473 |
1474 |9: // Handle stack expansion on return from yield.
1475 | mr CARG1, L
1476 | srwi CARG2, RD, 3
1477 | bl extern lj_state_growstack // (lua_State *L, int n)
1478 | li CRET1, 0
1479 | b <4
1480 |.endmacro
1481 |
1482 | coroutine_resume_wrap 1 // coroutine.resume
1483 | coroutine_resume_wrap 0 // coroutine.wrap
1484 |
1485 |.ffunc coroutine_yield
1486 | lwz TMP0, L->cframe
1487 | add TMP1, BASE, NARGS8:RC
1488 | stw BASE, L->base
1489 | andi. TMP0, TMP0, CFRAME_RESUME
1490 | stw TMP1, L->top
1491 | li CRET1, LUA_YIELD
1492 | beq ->fff_fallback
1493 | stw ZERO, L->cframe
1494 | stb CRET1, L->status
1495 | b ->vm_leave_unw
1496 |
1497 |//-- Math library -------------------------------------------------------
1498 |
1499 |.ffunc_1 math_abs
1500 | checknum CARG3
1501 if (LJ_DUALNUM) {
1502 | bne >2
1503 | srawi TMP1, CARG1, 31
1504 | xor TMP2, TMP1, CARG1
1505 | sub. CARG1, TMP2, TMP1
1506 | blt >1
1507 |->fff_resi:
1508 | lwz PC, FRAME_PC(BASE)
1509 | la RA, -8(BASE)
1510 | stw TISNUM, -8(BASE)
1511 | stw CRET1, -4(BASE)
1512 | b ->fff_res1
1513 |1:
1514 | lus CARG3, 0x41e0 // 2^31.
1515 | li CARG1, 0
1516 | b ->fff_restv
1517 |2:
1518 }
1519 | bge ->fff_fallback
1520 | rlwinm CARG3, CARG3, 0, 1, 31
1521 | // Fallthrough.
1522 |
1523 |->fff_restv:
1524 | // CARG3/CARG1 = TValue result.
1525 | lwz PC, FRAME_PC(BASE)
1526 | stw CARG3, -8(BASE)
1527 | la RA, -8(BASE)
1528 | stw CARG1, -4(BASE)
1529 |->fff_res1:
1530 | // RA = results, PC = return.
1531 | li RD, (1+1)*8
1532 |->fff_res:
1533 | // RA = results, RD = (nresults+1)*8, PC = return.
1534 | andi. TMP0, PC, FRAME_TYPE
1535 | mr MULTRES, RD
1536 | bney ->vm_return
1537 | lwz INS, -4(PC)
1538 | decode_RB8 RB, INS
1539 |5:
1540 | cmplw RB, RD // More results expected?
1541 | decode_RA8 TMP0, INS
1542 | bgt >6
1543 | ins_next1
1544 | // Adjust BASE. KBASE is assumed to be set for the calling frame.
1545 | sub BASE, RA, TMP0
1546 | ins_next2
1547 |
1548 |6: // Fill up results with nil.
1549 | subi TMP1, RD, 8
1550 | addi RD, RD, 8
1551 | stwx TISNIL, RA, TMP1
1552 | b <5
1553 |
1554 |.macro math_extern, func
1555 | .ffunc_n math_ .. func
1556 | bl extern func
1557 | b ->fff_resn
1558 |.endmacro
1559 |
1560 |.macro math_extern2, func
1561 | .ffunc_nn math_ .. func
1562 | bl extern func
1563 | b ->fff_resn
1564 |.endmacro
1565 |
1566 |.macro math_round, func
1567 | .ffunc_1 math_ .. func
1568 | checknum CARG3; beqy ->fff_restv
1569 | rlwinm TMP2, CARG3, 12, 21, 31
1570 | bge ->fff_fallback
1571 | addic. TMP2, TMP2, -1023 // exp = exponent(x) - 1023
1572 | cmplwi cr1, TMP2, 31 // 0 <= exp < 31?
1573 | subfic TMP0, TMP2, 31
1574 | blt >3
1575 | slwi TMP1, CARG3, 11
1576 | srwi TMP3, CARG1, 21
1577 | oris TMP1, TMP1, 0x8000
1578 | addi TMP2, TMP2, 1
1579 | or TMP1, TMP1, TMP3
1580 | slwi CARG2, CARG1, 11
1581 | bge cr1, >4
1582 | slw TMP3, TMP1, TMP2
1583 | srw CARG1, TMP1, TMP0
1584 | or TMP3, TMP3, CARG2
1585 | srawi TMP2, CARG3, 31
1586 |.if "func" == "floor"
1587 | and TMP1, TMP3, TMP2
1588 | addic TMP0, TMP1, -1
1589 | subfe TMP1, TMP0, TMP1
1590 | add CARG1, CARG1, TMP1
1591 | xor CARG1, CARG1, TMP2
1592 | sub CARG1, CARG1, TMP2
1593 | b ->fff_resi
1594 |.else
1595 | andc TMP1, TMP3, TMP2
1596 | addic TMP0, TMP1, -1
1597 | subfe TMP1, TMP0, TMP1
1598 | addo. CARG1, CARG1, TMP1
1599 | xor CARG1, CARG1, TMP2
1600 | sub CARG1, CARG1, TMP2
1601 | bns ->fff_resi
1602 | // Potential overflow.
1603 | mcrxr cr0; bley ->fff_resi // Ignore unrelated overflow.
1604 | lus CARG3, 0x41e0 // 2^31.
1605 | li CARG1, 0
1606 | b ->fff_restv
1607 |.endif
1608 |3: // |x| < 1
1609 | add TMP2, CARG3, CARG3
1610 | srawi TMP1, CARG3, 31
1611 | or TMP2, CARG1, TMP2 // ztest = (hi+hi) | lo
1612 |.if "func" == "floor"
1613 | and TMP1, TMP2, TMP1 // (ztest & sign) == 0 ? 0 : -1
1614 | subfic TMP2, TMP1, 0
1615 | subfe CARG1, CARG1, CARG1
1616 |.else
1617 | andc TMP1, TMP2, TMP1 // (ztest & ~sign) == 0 ? 0 : 1
1618 | addic TMP2, TMP1, -1
1619 | subfe CARG1, TMP2, TMP1
1620 |.endif
1621 | b ->fff_resi
1622 |4: // exp >= 31. Check for -(2^31).
1623 | xoris TMP1, TMP1, 0x8000
1624 | srawi TMP2, CARG3, 31
1625 |.if "func" == "floor"
1626 | or TMP1, TMP1, CARG2
1627 |.endif
1628 | orc. TMP1, TMP1, TMP2
1629 | crand 4*cr0+eq, 4*cr0+eq, 4*cr1+eq
1630 | lus CARG1, 0x8000 // -(2^31).
1631 | beqy ->fff_resi
1632 |5:
1633 | lfd FARG1, 0(BASE)
1634 | bl extern func
1635 | b ->fff_resn
1636 |.endmacro
1637 |
1638 if (LJ_DUALNUM) {
1639 | math_round floor
1640 | math_round ceil
1641 } else {
1642 | // NYI: use internal implementation.
1643 | math_extern floor
1644 | math_extern ceil
1645 }
1646 |
1647 | math_extern sqrt
1648 | math_extern log
1649 | math_extern log10
1650 | math_extern exp
1651 | math_extern sin
1652 | math_extern cos
1653 | math_extern tan
1654 | math_extern asin
1655 | math_extern acos
1656 | math_extern atan
1657 | math_extern sinh
1658 | math_extern cosh
1659 | math_extern tanh
1660 | math_extern2 pow
1661 | math_extern2 atan2
1662 | math_extern2 fmod
1663 |
1664 |->ff_math_deg:
1665 |.ffunc_n math_rad
1666 | lfd FARG2, CFUNC:RB->upvalue[0]
1667 | fmul FARG1, FARG1, FARG2
1668 | b ->fff_resn
1669 |
1670 if (LJ_DUALNUM) {
1671 |.ffunc math_ldexp
1672 | cmplwi NARGS8:RC, 16
1673 | lwz CARG3, 0(BASE)
1674 | lfd FARG1, 0(BASE)
1675 | lwz CARG4, 8(BASE)
1676 | lwz CARG1, 12(BASE)
1677 | blt ->fff_fallback
1678 | checknum CARG3; bge ->fff_fallback
1679 | checknum CARG4; bne ->fff_fallback
1680 } else {
1681 |.ffunc_nn math_ldexp
1682 | toint CARG1, FARG2
1683 }
1684 | bl extern ldexp
1685 | b ->fff_resn
1686 |
1687 |.ffunc_n math_frexp
1688 | la CARG1, DISPATCH_GL(tmptv)(DISPATCH)
1689 | lwz PC, FRAME_PC(BASE)
1690 | bl extern frexp
1691 | lwz TMP1, DISPATCH_GL(tmptv)(DISPATCH)
1692 | la RA, -8(BASE)
1693 if (!LJ_DUALNUM) {
1694 | tonum_i FARG2, TMP1
1695 }
1696 | stfd FARG1, 0(RA)
1697 | li RD, (2+1)*8
1698 if (LJ_DUALNUM) {
1699 | stw TISNUM, 8(RA)
1700 | stw TMP1, 12(RA)
1701 } else {
1702 | stfd FARG2, 8(RA)
1703 }
1704 | b ->fff_res
1705 |
1706 |.ffunc_n math_modf
1707 | la CARG1, -8(BASE)
1708 | lwz PC, FRAME_PC(BASE)
1709 | bl extern modf
1710 | la RA, -8(BASE)
1711 | stfd FARG1, 0(BASE)
1712 | li RD, (2+1)*8
1713 | b ->fff_res
1714 |
1715 |.macro math_minmax, name, ismax
1716 ||if (LJ_DUALNUM) {
1717 | .ffunc_1 name
1718 | checknum CARG3
1719 | addi TMP1, BASE, 8
1720 | add TMP2, BASE, NARGS8:RC
1721 | bne >4
1722 |1: // Handle integers.
1723 | lwz CARG4, 0(TMP1)
1724 | cmplw cr1, TMP1, TMP2
1725 | lwz CARG2, 4(TMP1)
1726 | bge cr1, ->fff_resi
1727 | checknum CARG4
1728 | xoris TMP0, CARG1, 0x8000
1729 | xoris TMP3, CARG2, 0x8000
1730 | bne >3
1731 | subfc TMP3, TMP3, TMP0
1732 | subfe TMP0, TMP0, TMP0
1733 |.if ismax
1734 | andc TMP3, TMP3, TMP0
1735 |.else
1736 | and TMP3, TMP3, TMP0
1737 |.endif
1738 | add CARG1, TMP3, CARG2
1739 | addi TMP1, TMP1, 8
1740 | b <1
1741 |3:
1742 | bge ->fff_fallback
1743 | // Convert intermediate result to number and continue below.
1744 | tonum_i FARG1, CARG1
1745 | lfd FARG2, 0(TMP1)
1746 | b >6
1747 |4:
1748 | lfd FARG1, 0(BASE)
1749 | bge ->fff_fallback
1750 |5: // Handle numbers.
1751 | lwz CARG4, 0(TMP1)
1752 | cmplw cr1, TMP1, TMP2
1753 | lfd FARG2, 0(TMP1)
1754 | bge cr1, ->fff_resn
1755 | checknum CARG4; bge >7
1756 |6:
1757 | fsub f0, FARG1, FARG2
1758 | addi TMP1, TMP1, 8
1759 |.if ismax
1760 | fsel FARG1, f0, FARG1, FARG2
1761 |.else
1762 | fsel FARG1, f0, FARG2, FARG1
1763 |.endif
1764 | b <5
1765 |7: // Convert integer to number and continue above.
1766 | lwz CARG2, 4(TMP1)
1767 | bne ->fff_fallback
1768 | tonum_i FARG2, CARG2
1769 | b <6
1770 ||} else {
1771 | .ffunc_n name
1772 | li TMP1, 8
1773 |1:
1774 | lwzx CARG2, BASE, TMP1
1775 | lfdx FARG2, BASE, TMP1
1776 | cmplw cr1, TMP1, NARGS8:RC
1777 | checknum CARG2
1778 | bge cr1, ->fff_resn
1779 | bge ->fff_fallback
1780 | fsub f0, FARG1, FARG2
1781 | addi TMP1, TMP1, 8
1782 |.if ismax
1783 | fsel FARG1, f0, FARG1, FARG2
1784 |.else
1785 | fsel FARG1, f0, FARG2, FARG1
1786 |.endif
1787 | b <1
1788 ||}
1789 |.endmacro
1790 |
1791 | math_minmax math_min, 0
1792 | math_minmax math_max, 1
1793 |
1794 |//-- String library -----------------------------------------------------
1795 |
1796 |.ffunc_1 string_len
1797 | checkstr CARG3; bne ->fff_fallback
1798 | lwz CRET1, STR:CARG1->len
1799 | b ->fff_resi
1800 |
1801 |.ffunc string_byte // Only handle the 1-arg case here.
1802 | cmplwi NARGS8:RC, 8
1803 | lwz CARG3, 0(BASE)
1804 | lwz STR:CARG1, 4(BASE)
1805 | bne ->fff_fallback // Need exactly 1 argument.
1806 | checkstr CARG3
1807 | bne ->fff_fallback
1808 | lwz TMP0, STR:CARG1->len
1809 if (LJ_DUALNUM) {
1810 | lbz CARG1, STR:CARG1[1] // Access is always ok (NUL at end).
1811 | li RD, (0+1)*8
1812 | lwz PC, FRAME_PC(BASE)
1813 | cmplwi TMP0, 0
1814 | la RA, -8(BASE)
1815 | beqy ->fff_res
1816 | b ->fff_resi
1817 } else {
1818 | lbz TMP1, STR:CARG1[1] // Access is always ok (NUL at end).
1819 | addic TMP3, TMP0, -1 // RD = ((str->len != 0)+1)*8
1820 | subfe RD, TMP3, TMP0
1821 | stw TMP1, TONUM_LO // Inlined tonum_u f0, TMP1.
1822 | addi RD, RD, 1
1823 | lfd f0, TONUM_D
1824 | la RA, -8(BASE)
1825 | lwz PC, FRAME_PC(BASE)
1826 | fsub f0, f0, TOBIT
1827 | slwi RD, RD, 3
1828 | stfd f0, 0(RA)
1829 | b ->fff_res
1830 }
1831 |
1832 |.ffunc string_char // Only handle the 1-arg case here.
1833 | ffgccheck
1834 | cmplwi NARGS8:RC, 8
1835 | lwz CARG3, 0(BASE)
1836 if (LJ_DUALNUM) {
1837 | lwz TMP0, 4(BASE)
1838 | bne ->fff_fallback // Exactly 1 argument.
1839 | checknum CARG3; bne ->fff_fallback
1840 | la CARG2, 7(BASE)
1841 } else {
1842 | lfd FARG1, 0(BASE)
1843 | bne ->fff_fallback // Exactly 1 argument.
1844 | checknum CARG3; bge ->fff_fallback
1845 | toint TMP0, FARG1
1846 | la CARG2, TMPD_BLO
1847 }
1848 | li CARG3, 1
1849 | cmplwi TMP0, 255; bgt ->fff_fallback
1850 |->fff_newstr:
1851 | mr CARG1, L
1852 | stw BASE, L->base
1853 | stw PC, SAVE_PC
1854 | bl extern lj_str_new // (lua_State *L, char *str, size_t l)
1855 | // Returns GCstr *.
1856 | lwz BASE, L->base
1857 | li CARG3, LJ_TSTR
1858 | b ->fff_restv
1859 |
1860 |.ffunc string_sub
1861 | ffgccheck
1862 | cmplwi NARGS8:RC, 16
1863 | lwz CARG3, 16(BASE)
1864 if (!LJ_DUALNUM) {
1865 | lfd f0, 16(BASE)
1866 }
1867 | lwz TMP0, 0(BASE)
1868 | lwz STR:CARG1, 4(BASE)
1869 | blt ->fff_fallback
1870 | lwz CARG2, 8(BASE)
1871 if (LJ_DUALNUM) {
1872 | lwz TMP1, 12(BASE)
1873 } else {
1874 | lfd f1, 8(BASE)
1875 }
1876 | li TMP2, -1
1877 | beq >1
1878 if (LJ_DUALNUM) {
1879 | checknum CARG3
1880 | lwz TMP2, 20(BASE)
1881 | bne ->fff_fallback
1882 |1:
1883 | checknum CARG2; bne ->fff_fallback
1884 } else {
1885 | checknum CARG3; bge ->fff_fallback
1886 | toint TMP2, f0
1887 |1:
1888 | checknum CARG2; bge ->fff_fallback
1889 }
1890 | checkstr TMP0; bne ->fff_fallback
1891 if (!LJ_DUALNUM) {
1892 | toint TMP1, f1
1893 }
1894 | lwz TMP0, STR:CARG1->len
1895 | cmplw TMP0, TMP2 // len < end? (unsigned compare)
1896 | addi TMP3, TMP2, 1
1897 | blt >5
1898 |2:
1899 | cmpwi TMP1, 0 // start <= 0?
1900 | add TMP3, TMP1, TMP0
1901 | ble >7
1902 |3:
1903 | sub CARG3, TMP2, TMP1
1904 | addi CARG2, STR:CARG1, #STR-1
1905 | srawi TMP0, CARG3, 31
1906 | addi CARG3, CARG3, 1
1907 | add CARG2, CARG2, TMP1
1908 | andc CARG3, CARG3, TMP0
1909 | b ->fff_newstr
1910 |
1911 |5: // Negative end or overflow.
1912 | sub CARG2, TMP0, TMP2
1913 | srawi CARG2, CARG2, 31
1914 | andc TMP3, TMP3, CARG2 // end = end > len ? len : end+len+1
1915 | add TMP2, TMP0, TMP3
1916 | b <2
1917 |
1918 |7: // Negative start or underflow.
1919 | addic CARG3, TMP1, -1
1920 | subfe CARG3, CARG3, CARG3
1921 | srawi CARG2, TMP3, 31 // Note: modifies carry.
1922 | andc TMP3, TMP3, CARG3
1923 | andc TMP1, TMP3, CARG2
1924 | addi TMP1, TMP1, 1 // start = 1 + (start ? start+len : 0)
1925 | b <3
1926 |
1927 |.ffunc string_rep // Only handle the 1-char case inline.
1928 | ffgccheck
1929 | cmplwi NARGS8:RC, 16
1930 | lwz TMP0, 0(BASE)
1931 | lwz STR:CARG1, 4(BASE)
1932 | lwz CARG4, 8(BASE)
1933 if (LJ_DUALNUM) {
1934 | lwz CARG3, 12(BASE)
1935 } else {
1936 | lfd FARG2, 8(BASE)
1937 }
1938 | blt ->fff_fallback
1939 | checkstr TMP0; bne ->fff_fallback
1940 if (LJ_DUALNUM) {
1941 | checknum CARG4; bne ->fff_fallback
1942 } else {
1943 | checknum CARG4; bge ->fff_fallback
1944 | toint CARG3, FARG2
1945 }
1946 | lwz TMP0, STR:CARG1->len
1947 | cmpwi CARG3, 0
1948 | lwz TMP1, DISPATCH_GL(tmpbuf.sz)(DISPATCH)
1949 | ble >2 // Count <= 0? (or non-int)
1950 | cmplwi TMP0, 1
1951 | subi TMP2, CARG3, 1
1952 | blt >2 // Zero length string?
1953 | cmplw cr1, TMP1, CARG3
1954 | bne ->fff_fallback // Fallback for > 1-char strings.
1955 | lbz TMP0, STR:CARG1[1]
1956 | lwz CARG2, DISPATCH_GL(tmpbuf.buf)(DISPATCH)
1957 | blt cr1, ->fff_fallback
1958 |1: // Fill buffer with char. Yes, this is suboptimal code (do you care?).
1959 | cmplwi TMP2, 0
1960 | stbx TMP0, CARG2, TMP2
1961 | subi TMP2, TMP2, 1
1962 | bne <1
1963 | b ->fff_newstr
1964 |2: // Return empty string.
1965 | la STR:CARG1, DISPATCH_GL(strempty)(DISPATCH)
1966 | li CARG3, LJ_TSTR
1967 | b ->fff_restv
1968 |
1969 |.ffunc string_reverse
1970 | ffgccheck
1971 | cmplwi NARGS8:RC, 8
1972 | lwz CARG3, 0(BASE)
1973 | lwz STR:CARG1, 4(BASE)
1974 | blt ->fff_fallback
1975 | checkstr CARG3
1976 | lwz TMP1, DISPATCH_GL(tmpbuf.sz)(DISPATCH)
1977 | bne ->fff_fallback
1978 | lwz CARG3, STR:CARG1->len
1979 | la CARG1, #STR(STR:CARG1)
1980 | lwz CARG2, DISPATCH_GL(tmpbuf.buf)(DISPATCH)
1981 | li TMP2, 0
1982 | cmplw TMP1, CARG3
1983 | subi TMP3, CARG3, 1
1984 | blt ->fff_fallback
1985 |1: // Reverse string copy.
1986 | cmpwi TMP3, 0
1987 | lbzx TMP1, CARG1, TMP2
1988 | blty ->fff_newstr
1989 | stbx TMP1, CARG2, TMP3
1990 | subi TMP3, TMP3, 1
1991 | addi TMP2, TMP2, 1
1992 | b <1
1993 |
1994 |.macro ffstring_case, name, lo
1995 | .ffunc name
1996 | ffgccheck
1997 | cmplwi NARGS8:RC, 8
1998 | lwz CARG3, 0(BASE)
1999 | lwz STR:CARG1, 4(BASE)
2000 | blt ->fff_fallback
2001 | checkstr CARG3
2002 | lwz TMP1, DISPATCH_GL(tmpbuf.sz)(DISPATCH)
2003 | bne ->fff_fallback
2004 | lwz CARG3, STR:CARG1->len
2005 | la CARG1, #STR(STR:CARG1)
2006 | lwz CARG2, DISPATCH_GL(tmpbuf.buf)(DISPATCH)
2007 | cmplw TMP1, CARG3
2008 | li TMP2, 0
2009 | blt ->fff_fallback
2010 |1: // ASCII case conversion.
2011 | cmplw TMP2, CARG3
2012 | lbzx TMP1, CARG1, TMP2
2013 | bgey ->fff_newstr
2014 | subi TMP0, TMP1, lo
2015 | xori TMP3, TMP1, 0x20
2016 | addic TMP0, TMP0, -26
2017 | subfe TMP3, TMP3, TMP3
2018 | andi. TMP3, TMP3, 0x20
2019 | xor TMP1, TMP1, TMP3
2020 | stbx TMP1, CARG2, TMP2
2021 | addi TMP2, TMP2, 1
2022 | b <1
2023 |.endmacro
2024 |
2025 |ffstring_case string_lower, 65
2026 |ffstring_case string_upper, 97
2027 |
2028 |//-- Table library ------------------------------------------------------
2029 |
2030 |.ffunc_1 table_getn
2031 | checktab CARG3; bne ->fff_fallback
2032 | bl extern lj_tab_len // (GCtab *t)
2033 | // Returns uint32_t (but less than 2^31).
2034 | b ->fff_resi
2035 |
2036 |//-- Bit library --------------------------------------------------------
2037 |
2038 |.macro .ffunc_bit, name
2039 ||if (LJ_DUALNUM) {
2040 | .ffunc_1 bit_..name
2041 | checknum CARG3; bnel ->fff_tobit_fb
2042 ||} else {
2043 | .ffunc_n bit_..name
2044 | fadd FARG1, FARG1, TOBIT
2045 | stfd FARG1, TMPD
2046 | lwz CARG1, TMPD_LO
2047 ||}
2048 |.endmacro
2049 |
2050 |.macro .ffunc_bit_op, name, ins
2051 | .ffunc_bit name
2052 | addi TMP1, BASE, 8
2053 | add TMP2, BASE, NARGS8:RC
2054 |1:
2055 | lwz CARG4, 0(TMP1)
2056 | cmplw cr1, TMP1, TMP2
2057 ||if (LJ_DUALNUM) {
2058 | lwz CARG2, 4(TMP1)
2059 ||} else {
2060 | lfd FARG1, 0(TMP1)
2061 ||}
2062 | bgey cr1, ->fff_resi
2063 | checknum CARG4
2064 ||if (LJ_DUALNUM) {
2065 | bnel ->fff_bitop_fb
2066 ||} else {
2067 | fadd FARG1, FARG1, TOBIT
2068 | bge ->fff_fallback
2069 | stfd FARG1, TMPD
2070 | lwz CARG2, TMPD_LO
2071 ||}
2072 | ins CARG1, CARG1, CARG2
2073 | addi TMP1, TMP1, 8
2074 | b <1
2075 |.endmacro
2076 |
2077 |.ffunc_bit_op band, and
2078 |.ffunc_bit_op bor, or
2079 |.ffunc_bit_op bxor, xor
2080 |
2081 |.ffunc_bit bswap
2082 | rotlwi TMP0, CARG1, 8
2083 | rlwimi TMP0, CARG1, 24, 0, 7
2084 | rlwimi TMP0, CARG1, 24, 16, 23
2085 | mr CRET1, TMP0
2086 | b ->fff_resi
2087 |
2088 |.ffunc_bit bnot
2089 | not CRET1, CARG1
2090 | b ->fff_resi
2091 |
2092 |.macro .ffunc_bit_sh, name, ins, shmod
2093 ||if (LJ_DUALNUM) {
2094 | .ffunc_2 bit_..name
2095 | checknum CARG3; bnel ->fff_tobit_fb
2096 | // Note: no inline conversion from number for 2nd argument!
2097 | checknum CARG4; bne ->fff_fallback
2098 ||} else {
2099 | .ffunc_nn bit_..name
2100 | fadd FARG1, FARG1, TOBIT
2101 | fadd FARG2, FARG2, TOBIT
2102 | stfd FARG1, TMPD
2103 | lwz CARG1, TMPD_LO
2104 | stfd FARG2, TMPD
2105 | lwz CARG2, TMPD_LO
2106 ||}
2107 |.if shmod == 1
2108 | rlwinm CARG2, CARG2, 0, 27, 31
2109 |.elif shmod == 2
2110 | neg CARG2, CARG2
2111 |.endif
2112 | ins CRET1, CARG1, CARG2
2113 | b ->fff_resi
2114 |.endmacro
2115 |
2116 |.ffunc_bit_sh lshift, slw, 1
2117 |.ffunc_bit_sh rshift, srw, 1
2118 |.ffunc_bit_sh arshift, sraw, 1
2119 |.ffunc_bit_sh rol, rotlw, 0
2120 |.ffunc_bit_sh ror, rotlw, 2
2121 |
2122 |.ffunc_bit tobit
2123 if (LJ_DUALNUM) {
2124 | b ->fff_resi
2125 } else {
2126 |->fff_resi:
2127 | tonum_i FARG1, CRET1
2128 }
2129 |->fff_resn:
2130 | lwz PC, FRAME_PC(BASE)
2131 | la RA, -8(BASE)
2132 | stfd FARG1, -8(BASE)
2133 | b ->fff_res1
2134 |
2135 |// Fallback FP number to bit conversion.
2136 |->fff_tobit_fb:
2137 if (LJ_DUALNUM) {
2138 | lfd FARG1, 0(BASE)
2139 | bgt ->fff_fallback
2140 | fadd FARG1, FARG1, TOBIT
2141 | stfd FARG1, TMPD
2142 | lwz CARG1, TMPD_LO
2143 | blr
2144 }
2145 |->fff_bitop_fb:
2146 if (LJ_DUALNUM) {
2147 | lfd FARG1, 0(TMP1)
2148 | bgt ->fff_fallback
2149 | fadd FARG1, FARG1, TOBIT
2150 | stfd FARG1, TMPD
2151 | lwz CARG2, TMPD_LO
2152 | blr
2153 }
2154 |
2155 |//-----------------------------------------------------------------------
2156 |
2157 |->fff_fallback: // Call fast function fallback handler.
2158 | // BASE = new base, RB = CFUNC, RC = nargs*8
2159 | lwz TMP3, CFUNC:RB->f
2160 | add TMP1, BASE, NARGS8:RC
2161 | lwz PC, FRAME_PC(BASE) // Fallback may overwrite PC.
2162 | addi TMP0, TMP1, 8*LUA_MINSTACK
2163 | lwz TMP2, L->maxstack
2164 | stw PC, SAVE_PC // Redundant (but a defined value).
2165 | cmplw TMP0, TMP2
2166 | stw BASE, L->base
2167 | stw TMP1, L->top
2168 | mr CARG1, L
2169 | bgt >5 // Need to grow stack.
2170 | mtctr TMP3
2171 | bctrl // (lua_State *L)
2172 | // Either throws an error, or recovers and returns -1, 0 or nresults+1.
2173 | lwz BASE, L->base
2174 | cmpwi CRET1, 0
2175 | slwi RD, CRET1, 3
2176 | la RA, -8(BASE)
2177 | bgt ->fff_res // Returned nresults+1?
2178 |1: // Returned 0 or -1: retry fast path.
2179 | lwz TMP0, L->top
2180 | lwz LFUNC:RB, FRAME_FUNC(BASE)
2181 | sub NARGS8:RC, TMP0, BASE
2182 | bne ->vm_call_tail // Returned -1?
2183 | ins_callt // Returned 0: retry fast path.
2184 |
2185 |// Reconstruct previous base for vmeta_call during tailcall.
2186 |->vm_call_tail:
2187 | andi. TMP0, PC, FRAME_TYPE
2188 | rlwinm TMP1, PC, 0, 0, 28
2189 | bne >3
2190 | lwz INS, -4(PC)
2191 | decode_RA8 TMP1, INS
2192 | addi TMP1, TMP1, 8
2193 |3:
2194 | sub TMP2, BASE, TMP1
2195 | b ->vm_call_dispatch // Resolve again for tailcall.
2196 |
2197 |5: // Grow stack for fallback handler.
2198 | li CARG2, LUA_MINSTACK
2199 | bl extern lj_state_growstack // (lua_State *L, int n)
2200 | lwz BASE, L->base
2201 | cmpw TMP0, TMP0 // Set 4*cr0+eq to force retry.
2202 | b <1
2203 |
2204 |->fff_gcstep: // Call GC step function.
2205 | // BASE = new base, RC = nargs*8
2206 | mflr SAVE0
2207 | stw BASE, L->base
2208 | add TMP0, BASE, NARGS8:RC
2209 | stw PC, SAVE_PC // Redundant (but a defined value).
2210 | stw TMP0, L->top
2211 | mr CARG1, L
2212 | bl extern lj_gc_step // (lua_State *L)
2213 | lwz BASE, L->base
2214 | mtlr SAVE0
2215 | lwz TMP0, L->top
2216 | sub NARGS8:RC, TMP0, BASE
2217 | lwz CFUNC:RB, FRAME_FUNC(BASE)
2218 | blr
2219 |
2220 |//-----------------------------------------------------------------------
2221 |//-- Special dispatch targets -------------------------------------------
2222 |//-----------------------------------------------------------------------
2223 |
2224 |->vm_record: // Dispatch target for recording phase.
2225#if LJ_HASJIT
2226 | lbz TMP3, DISPATCH_GL(hookmask)(DISPATCH)
2227 | andi. TMP0, TMP3, HOOK_VMEVENT // No recording while in vmevent.
2228 | bne >5
2229 | // Decrement the hookcount for consistency, but always do the call.
2230 | lwz TMP2, DISPATCH_GL(hookcount)(DISPATCH)
2231 | andi. TMP0, TMP3, HOOK_ACTIVE
2232 | bne >1
2233 | subi TMP2, TMP2, 1
2234 | andi. TMP0, TMP3, LUA_MASKLINE|LUA_MASKCOUNT
2235 | beqy >1
2236 | stw TMP2, DISPATCH_GL(hookcount)(DISPATCH)
2237 | b >1
2238#endif
2239 |
2240 |->vm_rethook: // Dispatch target for return hooks.
2241 | lbz TMP3, DISPATCH_GL(hookmask)(DISPATCH)
2242 | andi. TMP0, TMP3, HOOK_ACTIVE // Hook already active?
2243 | beq >1
2244 |5: // Re-dispatch to static ins.
2245 | addi TMP1, TMP1, GG_DISP2STATIC // Assumes decode_OP4 TMP1, INS.
2246 | lwzx TMP0, DISPATCH, TMP1
2247 | mtctr TMP0
2248 | bctr
2249 |
2250 |->vm_inshook: // Dispatch target for instr/line hooks.
2251 | lbz TMP3, DISPATCH_GL(hookmask)(DISPATCH)
2252 | lwz TMP2, DISPATCH_GL(hookcount)(DISPATCH)
2253 | andi. TMP0, TMP3, HOOK_ACTIVE // Hook already active?
2254 | rlwinm TMP0, TMP3, 31-LUA_HOOKLINE, 31, 0
2255 | bne <5
2256 |
2257 | cmpwi cr1, TMP0, 0
2258 | addic. TMP2, TMP2, -1
2259 | beq cr1, <5
2260 | stw TMP2, DISPATCH_GL(hookcount)(DISPATCH)
2261 | beq >1
2262 | bge cr1, <5
2263 |1:
2264 | mr CARG1, L
2265 | stw MULTRES, SAVE_MULTRES
2266 | mr CARG2, PC
2267 | stw BASE, L->base
2268 | // SAVE_PC must hold the _previous_ PC. The callee updates it with PC.
2269 | bl extern lj_dispatch_ins // (lua_State *L, const BCIns *pc)
2270 |3:
2271 | lwz BASE, L->base
2272 |4: // Re-dispatch to static ins.
2273 | lwz INS, -4(PC)
2274 | decode_OP4 TMP1, INS
2275 | decode_RB8 RB, INS
2276 | addi TMP1, TMP1, GG_DISP2STATIC
2277 | decode_RD8 RD, INS
2278 | lwzx TMP0, DISPATCH, TMP1
2279 | decode_RA8 RA, INS
2280 | decode_RC8 RC, INS
2281 | mtctr TMP0
2282 | bctr
2283 |
2284 |->cont_hook: // Continue from hook yield.
2285 | addi PC, PC, 4
2286 | lwz MULTRES, -20(RB) // Restore MULTRES for *M ins.
2287 | b <4
2288 |
2289 |->vm_hotloop: // Hot loop counter underflow.
2290#if LJ_HASJIT
2291 | lwz LFUNC:TMP1, FRAME_FUNC(BASE)
2292 | addi CARG1, DISPATCH, GG_DISP2J
2293 | stw PC, SAVE_PC
2294 | lwz TMP1, LFUNC:TMP1->pc
2295 | mr CARG2, PC
2296 | stw L, DISPATCH_J(L)(DISPATCH)
2297 | lbz TMP1, PC2PROTO(framesize)(TMP1)
2298 | stw BASE, L->base
2299 | slwi TMP1, TMP1, 3
2300 | add TMP1, BASE, TMP1
2301 | stw TMP1, L->top
2302 | bl extern lj_trace_hot // (jit_State *J, const BCIns *pc)
2303 | b <3
2304#endif
2305 |
2306 |->vm_callhook: // Dispatch target for call hooks.
2307 | mr CARG2, PC
2308#if LJ_HASJIT
2309 | b >1
2310#endif
2311 |
2312 |->vm_hotcall: // Hot call counter underflow.
2313#if LJ_HASJIT
2314 | ori CARG2, PC, 1
2315 |1:
2316#endif
2317 | add TMP0, BASE, RC
2318 | stw PC, SAVE_PC
2319 | mr CARG1, L
2320 | stw BASE, L->base
2321 | sub RA, RA, BASE
2322 | stw TMP0, L->top
2323 | bl extern lj_dispatch_call // (lua_State *L, const BCIns *pc)
2324 | // Returns ASMFunction.
2325 | lwz BASE, L->base
2326 | lwz TMP0, L->top
2327 | stw ZERO, SAVE_PC // Invalidate for subsequent line hook.
2328 | sub NARGS8:RC, TMP0, BASE
2329 | add RA, BASE, RA
2330 | lwz LFUNC:RB, FRAME_FUNC(BASE)
2331 | lwz INS, -4(PC)
2332 | mtctr CRET1
2333 | bctr
2334 |
2335 |//-----------------------------------------------------------------------
2336 |//-- Trace exit handler -------------------------------------------------
2337 |//-----------------------------------------------------------------------
2338 |
2339 |.macro savex_, a, b, c, d
2340 | stfd f..a, 16+a*8(sp)
2341 | stfd f..b, 16+b*8(sp)
2342 | stfd f..c, 16+c*8(sp)
2343 | stfd f..d, 16+d*8(sp)
2344 |.endmacro
2345 |
2346 |->vm_exit_handler:
2347#if LJ_HASJIT
2348 | addi sp, sp, -(16+32*8+32*4)
2349 | stmw r2, 16+32*8+2*4(sp)
2350 | addi DISPATCH, JGL, -GG_DISP2G-32768
2351 | li CARG2, ~LJ_VMST_EXIT
2352 | lwz CARG1, 16+32*8+32*4(sp) // Get stack chain.
2353 | stw CARG2, DISPATCH_GL(vmstate)(DISPATCH)
2354 | savex_ 0,1,2,3
2355 | stw CARG1, 0(sp) // Store extended stack chain.
2356 | mcrxr cr0 // Clear SO flag.
2357 | savex_ 4,5,6,7
2358 | addi CARG2, sp, 16+32*8+32*4 // Recompute original value of sp.
2359 | savex_ 8,9,10,11
2360 | stw CARG2, 16+32*8+1*4(sp) // Store sp in RID_SP.
2361 | savex_ 12,13,14,15
2362 | mflr CARG3
2363 | li TMP1, 0
2364 | savex_ 16,17,18,19
2365 | stw TMP1, 16+32*8+0*4(sp) // Clear RID_TMP.
2366 | savex_ 20,21,22,23
2367 | lhz CARG4, 2(CARG3) // Load trace number.
2368 | savex_ 24,25,26,27
2369 | lwz L, DISPATCH_GL(jit_L)(DISPATCH)
2370 | savex_ 28,29,30,31
2371 | sub CARG3, TMP0, CARG3 // Compute exit number.
2372 | lwz BASE, DISPATCH_GL(jit_base)(DISPATCH)
2373 | srwi CARG3, CARG3, 2
2374 | stw L, DISPATCH_J(L)(DISPATCH)
2375 | subi CARG3, CARG3, 2
2376 | stw TMP1, DISPATCH_GL(jit_L)(DISPATCH)
2377 | stw CARG4, DISPATCH_J(parent)(DISPATCH)
2378 | stw BASE, L->base
2379 | addi CARG1, DISPATCH, GG_DISP2J
2380 | stw CARG3, DISPATCH_J(exitno)(DISPATCH)
2381 | addi CARG2, sp, 16
2382 | bl extern lj_trace_exit // (jit_State *J, ExitState *ex)
2383 | // Returns MULTRES (unscaled) or negated error code.
2384 | lwz TMP1, L->cframe
2385 | lwz TMP2, 0(sp)
2386 | lwz BASE, L->base
2387 | rlwinm sp, TMP1, 0, 0, 29
2388 | lwz PC, SAVE_PC // Get SAVE_PC.
2389 | stw TMP2, 0(sp)
2390 | stw L, SAVE_L // Set SAVE_L (on-trace resume/yield).
2391 | b >1
2392#endif
2393 |->vm_exit_interp:
2394#if LJ_HASJIT
2395 | // CARG1 = MULTRES or negated error code, BASE, PC and JGL set.
2396 | lwz L, SAVE_L
2397 | addi DISPATCH, JGL, -GG_DISP2G-32768
2398 |1:
2399 | cmpwi CARG1, 0
2400 | blt >3 // Check for error from exit.
2401 | lwz LFUNC:TMP1, FRAME_FUNC(BASE)
2402 | slwi MULTRES, CARG1, 3
2403 | li TMP2, 0
2404 | stw MULTRES, SAVE_MULTRES
2405 | lwz TMP1, LFUNC:TMP1->pc
2406 | stw TMP2, DISPATCH_GL(jit_L)(DISPATCH)
2407 | lwz KBASE, PC2PROTO(k)(TMP1)
2408 | // Setup type comparison constants.
2409 | li TISNUM, LJ_TISNUM
2410 | lus TMP3, 0x59c0 // TOBIT = 2^52 + 2^51 (float).
2411 | stw TMP3, TMPD
2412 | li ZERO, 0
2413 | ori TMP3, TMP3, 0x0004 // TONUM = 2^52 + 2^51 + 2^31 (float).
2414 | lfs TOBIT, TMPD
2415 | stw TMP3, TMPD
2416 | lus TMP0, 0x4338 // Hiword of 2^52 + 2^51 (double)
2417 | li TISNIL, LJ_TNIL
2418 | stw TMP0, TONUM_HI
2419 | lfs TONUM, TMPD
2420 | // Modified copy of ins_next which handles function header dispatch, too.
2421 | lwz INS, 0(PC)
2422 | addi PC, PC, 4
2423 | // Assumes TISNIL == ~LJ_VMST_INTERP == -1.
2424 | stw TISNIL, DISPATCH_GL(vmstate)(DISPATCH)
2425 | decode_OP4 TMP1, INS
2426 | decode_RA8 RA, INS
2427 | lwzx TMP0, DISPATCH, TMP1
2428 | mtctr TMP0
2429 | cmplwi TMP1, BC_FUNCF*4 // Function header?
2430 | bge >2
2431 | decode_RB8 RB, INS
2432 | decode_RD8 RD, INS
2433 | decode_RC8 RC, INS
2434 | bctr
2435 |2:
2436 | subi RC, MULTRES, 8
2437 | add RA, RA, BASE
2438 | bctr
2439 |
2440 |3: // Rethrow error from the right C frame.
2441 | neg CARG2, CARG1
2442 | mr CARG1, L
2443 | bl extern lj_err_throw // (lua_State *L, int errcode)
2444#endif
2445 |
2446 |//-----------------------------------------------------------------------
2447 |//-- Math helper functions ----------------------------------------------
2448 |//-----------------------------------------------------------------------
2449 |
2450 | // NYI: Use internal implementation.
2451 |->vm_floor:
2452 | b extern floor
2453 |->vm_ceil:
2454 | b extern ceil
2455 |->vm_trunc:
2456#if LJ_HASJIT
2457 | b extern trunc
2458#endif
2459 |
2460 |->vm_modi:
2461 | divwo. TMP0, CARG1, CARG2
2462 | bso >1
2463 | xor. CARG3, CARG1, CARG2
2464 | mullw TMP0, TMP0, CARG2
2465 | sub CARG1, CARG1, TMP0
2466 | bgelr
2467 | cmpwi CARG1, 0; beqlr
2468 | add CARG1, CARG1, CARG2
2469 | blr
2470 |1:
2471 | cmpwi CARG2, 0
2472 | li CARG1, 0
2473 | beqlr
2474 | mcrxr cr0 // Clear SO for -2147483648 % -1 and return 0.
2475 | blr
2476 |
2477 |// Callable from C: double lj_vm_foldarith(double x, double y, int op)
2478 |// Compute x op y for basic arithmetic operators (+ - * / % ^ and unary -)
2479 |// and basic math functions. ORDER ARITH
2480 |->vm_foldarith:
2481 | cmplwi CARG1, 1
2482 | beq >1; bgt >2
2483 | fadd FARG1, FARG1, FARG2; blr
2484 |1:
2485 | fsub FARG1, FARG1, FARG2; blr
2486 |2:
2487 | cmplwi CARG1, 3; beq >1; bgt >2
2488 | fmul FARG1, FARG1, FARG2; blr
2489 |1:
2490 | fdiv FARG1, FARG1, FARG2; blr
2491 |2:
2492 | cmplwi CARG1, 5; beq >1; bgt >2
2493 | // NYI: Use internal implementation of floor and avoid spills.
2494 | stwu sp, -32(sp); stfd f14, 16(sp); stfd f15, 24(sp)
2495 | mflr r0
2496 | fmr f14, FARG1
2497 | fdiv FARG1, FARG1, FARG2
2498 | stw r0, 36(sp)
2499 | fmr f15, FARG2
2500 | bl extern floor
2501 | lwz r0, 36(sp)
2502 | fmul FARG1, FARG1, f15
2503 | mtlr r0
2504 | fsub FARG1, f14, FARG1
2505 | lfd f14, 16(sp); lfd f15, 24(sp); addi sp, sp, 32; blr
2506 |1:
2507 | b extern pow
2508 |2:
2509 | cmplwi CARG1, 7; beq >1; bgt >2
2510 | fneg FARG1, FARG1; blr
2511 |1:
2512 | fabs FARG1, FARG1; blr
2513 |2:
2514#if LJ_HASJIT
2515 | cmplwi CARG1, 9; beq >9; bgt >2
2516 | b extern atan2
2517 | // No support needed for IR_LDEXP.
2518 |2:
2519 | cmplwi CARG1, 11; bgt >9
2520 | fsub f0, FARG1, FARG2
2521 | beq >1
2522 | fsel FARG1, f0, FARG2, FARG1 // IR_MAX
2523 | blr
2524 |1:
2525 | fsel FARG1, f0, FARG1, FARG2 // IR_MIN
2526 | blr
2527 |9:
2528 | NYI // Bad op.
2529#else
2530 | NYI // Other operations only needed by JIT compiler.
2531#endif
2532 |
2533 |//-----------------------------------------------------------------------
2534 |//-- Miscellaneous functions --------------------------------------------
2535 |//-----------------------------------------------------------------------
2536 |
2537 |// void lj_vm_cachesync(void *start, void *end)
2538 |// Flush D-Cache and invalidate I-Cache. Assumes 32 byte cache line size.
2539 |// This is a good lower bound, except for very ancient PPC models.
2540 |->vm_cachesync:
2541 | // Compute start of first cache line and number of cache lines.
2542 | rlwinm CARG1, CARG1, 0, 0, 26
2543 | sub CARG2, CARG2, CARG1
2544 | addi CARG2, CARG2, 31
2545 | rlwinm. CARG2, CARG2, 27, 5, 31
2546 | beqlr
2547 | mtctr CARG2
2548 | mr CARG3, CARG1
2549 |1: // Flush D-Cache.
2550 | dcbst r0, CARG1
2551 | addi CARG1, CARG1, 32
2552 | bdnz <1
2553 | sync
2554 | mtctr CARG2
2555 |1: // Invalidate I-Cache.
2556 | icbi r0, CARG3
2557 | addi CARG3, CARG3, 32
2558 | bdnz <1
2559 | isync
2560 | blr
2561 |
2562 |//-----------------------------------------------------------------------
2563 |//-- FFI helper functions -----------------------------------------------
2564 |//-----------------------------------------------------------------------
2565 |
2566 |// Handler for callback functions. Callback slot number in r11, g in r12.
2567 |->vm_ffi_callback:
2568#if LJ_HASFFI
2569 |.type CTSTATE, CTState, PC
2570 | saveregs
2571 | lwz CTSTATE, GL:r12->ctype_state
2572 | addi DISPATCH, r12, GG_G2DISP
2573 | stw r11, CTSTATE->cb.slot
2574 | stw r3, CTSTATE->cb.gpr[0]
2575 | stfd f1, CTSTATE->cb.fpr[0]
2576 | stw r4, CTSTATE->cb.gpr[1]
2577 | stfd f2, CTSTATE->cb.fpr[1]
2578 | stw r5, CTSTATE->cb.gpr[2]
2579 | stfd f3, CTSTATE->cb.fpr[2]
2580 | stw r6, CTSTATE->cb.gpr[3]
2581 | stfd f4, CTSTATE->cb.fpr[3]
2582 | stw r7, CTSTATE->cb.gpr[4]
2583 | stfd f5, CTSTATE->cb.fpr[4]
2584 | stw r8, CTSTATE->cb.gpr[5]
2585 | stfd f6, CTSTATE->cb.fpr[5]
2586 | stw r9, CTSTATE->cb.gpr[6]
2587 | stfd f7, CTSTATE->cb.fpr[6]
2588 | stw r10, CTSTATE->cb.gpr[7]
2589 | stfd f8, CTSTATE->cb.fpr[7]
2590 | addi TMP0, sp, CFRAME_SPACE+8
2591 | stw TMP0, CTSTATE->cb.stack
2592 | mr CARG1, CTSTATE
2593 | stw CTSTATE, SAVE_PC // Any value outside of bytecode is ok.
2594 | mr CARG2, sp
2595 | bl extern lj_ccallback_enter // (CTState *cts, void *cf)
2596 | // Returns lua_State *.
2597 | lwz BASE, L:CRET1->base
2598 | li TISNUM, LJ_TISNUM // Setup type comparison constants.
2599 | lwz RC, L:CRET1->top
2600 | lus TMP3, 0x59c0 // TOBIT = 2^52 + 2^51 (float).
2601 | li ZERO, 0
2602 | mr L, CRET1
2603 | stw TMP3, TMPD
2604 | lwz LFUNC:RB, FRAME_FUNC(BASE)
2605 | ori TMP3, TMP3, 0x0004 // TONUM = 2^52 + 2^51 + 2^31 (float).
2606 | li TISNIL, LJ_TNIL
2607 | li_vmstate INTERP
2608 | lfs TOBIT, TMPD
2609 | stw TMP3, TMPD
2610 | sub RC, RC, BASE
2611 | st_vmstate
2612 | lfs TONUM, TMPD
2613 | ins_callt
2614#endif
2615 |
2616 |->cont_ffi_callback: // Return from FFI callback.
2617#if LJ_HASFFI
2618 | lwz CTSTATE, DISPATCH_GL(ctype_state)(DISPATCH)
2619 | stw BASE, L->base
2620 | stw RB, L->top
2621 | stw L, CTSTATE->L
2622 | mr CARG1, CTSTATE
2623 | mr CARG2, RA
2624 | bl extern lj_ccallback_leave // (CTState *cts, TValue *o)
2625 | lwz CRET1, CTSTATE->cb.gpr[0]
2626 | lfd FARG1, CTSTATE->cb.fpr[0]
2627 | lwz CRET2, CTSTATE->cb.gpr[1]
2628 | b ->vm_leave_unw
2629#endif
2630 |
2631 |->vm_ffi_call: // Call C function via FFI.
2632 | // Caveat: needs special frame unwinding, see below.
2633#if LJ_HASFFI
2634 | .type CCSTATE, CCallState, CARG1
2635 | lwz TMP1, CCSTATE->spadj
2636 | mflr TMP0
2637 | lbz CARG2, CCSTATE->nsp
2638 | lbz CARG3, CCSTATE->nfpr
2639 | neg TMP1, TMP1
2640 | stw TMP0, 4(sp)
2641 | cmpwi cr1, CARG3, 0
2642 | mr TMP2, sp
2643 | addic. CARG2, CARG2, -1
2644 | stwux sp, sp, TMP1
2645 | crnot 4*cr1+eq, 4*cr1+eq // For vararg calls.
2646 | stw r14, -4(TMP2)
2647 | stw CCSTATE, -8(TMP2)
2648 | mr r14, TMP2
2649 | la TMP1, CCSTATE->stack
2650 | slwi CARG2, CARG2, 2
2651 | blty >2
2652 | la TMP2, 8(sp)
2653 |1:
2654 | lwzx TMP0, TMP1, CARG2
2655 | stwx TMP0, TMP2, CARG2
2656 | addic. CARG2, CARG2, -4
2657 | bge <1
2658 |2:
2659 | bney cr1, >3
2660 | lfd f1, CCSTATE->fpr[0]
2661 | lfd f2, CCSTATE->fpr[1]
2662 | lfd f3, CCSTATE->fpr[2]
2663 | lfd f4, CCSTATE->fpr[3]
2664 | lfd f5, CCSTATE->fpr[4]
2665 | lfd f6, CCSTATE->fpr[5]
2666 | lfd f7, CCSTATE->fpr[6]
2667 | lfd f8, CCSTATE->fpr[7]
2668 |3:
2669 | lwz TMP0, CCSTATE->func
2670 | lwz CARG2, CCSTATE->gpr[1]
2671 | lwz CARG3, CCSTATE->gpr[2]
2672 | lwz CARG4, CCSTATE->gpr[3]
2673 | lwz CARG5, CCSTATE->gpr[4]
2674 | mtctr TMP0
2675 | lwz r8, CCSTATE->gpr[5]
2676 | lwz r9, CCSTATE->gpr[6]
2677 | lwz r10, CCSTATE->gpr[7]
2678 | lwz CARG1, CCSTATE->gpr[0] // Do this last, since CCSTATE is CARG1.
2679 | bctrl
2680 | lwz CCSTATE:TMP1, -8(r14)
2681 | lwz TMP2, -4(r14)
2682 | lwz TMP0, 4(r14)
2683 | stw CARG1, CCSTATE:TMP1->gpr[0]
2684 | stfd FARG1, CCSTATE:TMP1->fpr[0]
2685 | stw CARG2, CCSTATE:TMP1->gpr[1]
2686 | mtlr TMP0
2687 | stw CARG3, CCSTATE:TMP1->gpr[2]
2688 | mr sp, r14
2689 | stw CARG4, CCSTATE:TMP1->gpr[3]
2690 | mr r14, TMP2
2691 | blr
2692#endif
2693 |// Note: vm_ffi_call must be the last function in this object file!
2694 |
2695 |//-----------------------------------------------------------------------
2696}
2697
2698/* Generate the code for a single instruction. */
2699static void build_ins(BuildCtx *ctx, BCOp op, int defop)
2700{
2701 int vk = 0;
2702 |=>defop:
2703
2704 switch (op) {
2705
2706 /* -- Comparison ops ---------------------------------------------------- */
2707
2708 /* Remember: all ops branch for a true comparison, fall through otherwise. */
2709
2710 case BC_ISLT: case BC_ISGE: case BC_ISLE: case BC_ISGT:
2711 | // RA = src1*8, RD = src2*8, JMP with RD = target
2712 if (LJ_DUALNUM) {
2713 | lwzux TMP0, RA, BASE
2714 | addi PC, PC, 4
2715 | lwz CARG2, 4(RA)
2716 | lwzux TMP1, RD, BASE
2717 | lwz TMP2, -4(PC)
2718 | checknum cr0, TMP0
2719 | lwz CARG3, 4(RD)
2720 | decode_RD4 TMP2, TMP2
2721 | checknum cr1, TMP1
2722 | addis TMP2, TMP2, -(BCBIAS_J*4 >> 16)
2723 | bne cr0, >7
2724 | bne cr1, >8
2725 | cmpw CARG2, CARG3
2726 if (op == BC_ISLT) {
2727 | bge >2
2728 } else if (op == BC_ISGE) {
2729 | blt >2
2730 } else if (op == BC_ISLE) {
2731 | bgt >2
2732 } else {
2733 | ble >2
2734 }
2735 |1:
2736 | add PC, PC, TMP2
2737 |2:
2738 | ins_next
2739 |
2740 |7: // RA is not an integer.
2741 | bgt cr0, ->vmeta_comp
2742 | // RA is a number.
2743 | lfd f0, 0(RA)
2744 | bgt cr1, ->vmeta_comp
2745 | blt cr1, >4
2746 | // RA is a number, RD is an integer.
2747 | tonum_i f1, CARG3
2748 | b >5
2749 |
2750 |8: // RA is an integer, RD is not an integer.
2751 | bgt cr1, ->vmeta_comp
2752 | // RA is an integer, RD is a number.
2753 | tonum_i f0, CARG2
2754 |4:
2755 | lfd f1, 0(RD)
2756 |5:
2757 | fcmpu cr0, f0, f1
2758 if (op == BC_ISLT) {
2759 | bge <2
2760 } else if (op == BC_ISGE) {
2761 | blt <2
2762 } else if (op == BC_ISLE) {
2763 | cror 4*cr0+lt, 4*cr0+lt, 4*cr0+eq
2764 | bge <2
2765 } else {
2766 | cror 4*cr0+lt, 4*cr0+lt, 4*cr0+eq
2767 | blt <2
2768 }
2769 | b <1
2770 } else {
2771 | lwzx TMP0, BASE, RA
2772 | addi PC, PC, 4
2773 | lfdx f0, BASE, RA
2774 | lwzx TMP1, BASE, RD
2775 | checknum cr0, TMP0
2776 | lwz TMP2, -4(PC)
2777 | lfdx f1, BASE, RD
2778 | checknum cr1, TMP1
2779 | decode_RD4 TMP2, TMP2
2780 | bge cr0, ->vmeta_comp
2781 | addis TMP2, TMP2, -(BCBIAS_J*4 >> 16)
2782 | bge cr1, ->vmeta_comp
2783 | fcmpu cr0, f0, f1
2784 if (op == BC_ISLT) {
2785 | bge >1
2786 } else if (op == BC_ISGE) {
2787 | blt >1
2788 } else if (op == BC_ISLE) {
2789 | cror 4*cr0+lt, 4*cr0+lt, 4*cr0+eq
2790 | bge >1
2791 } else {
2792 | cror 4*cr0+lt, 4*cr0+lt, 4*cr0+eq
2793 | blt >1
2794 }
2795 | add PC, PC, TMP2
2796 |1:
2797 | ins_next
2798 }
2799 break;
2800
2801 case BC_ISEQV: case BC_ISNEV:
2802 vk = op == BC_ISEQV;
2803 | // RA = src1*8, RD = src2*8, JMP with RD = target
2804 if (LJ_DUALNUM) {
2805 | lwzux TMP0, RA, BASE
2806 | addi PC, PC, 4
2807 | lwz CARG2, 4(RA)
2808 | lwzux TMP1, RD, BASE
2809 | checknum cr0, TMP0
2810 | lwz TMP2, -4(PC)
2811 | checknum cr1, TMP1
2812 | decode_RD4 TMP2, TMP2
2813 | lwz CARG3, 4(RD)
2814 | cror 4*cr7+gt, 4*cr0+gt, 4*cr1+gt
2815 | addis TMP2, TMP2, -(BCBIAS_J*4 >> 16)
2816 if (vk) {
2817 | ble cr7, ->BC_ISEQN_Z
2818 } else {
2819 | ble cr7, ->BC_ISNEN_Z
2820 }
2821 } else {
2822 | lwzux TMP0, RA, BASE
2823 | lwz TMP2, 0(PC)
2824 | lfd f0, 0(RA)
2825 | addi PC, PC, 4
2826 | lwzux TMP1, RD, BASE
2827 | checknum cr0, TMP0
2828 | decode_RD4 TMP2, TMP2
2829 | lfd f1, 0(RD)
2830 | checknum cr1, TMP1
2831 | addis TMP2, TMP2, -(BCBIAS_J*4 >> 16)
2832 | bge cr0, >5
2833 | bge cr1, >5
2834 | fcmpu cr0, f0, f1
2835 if (vk) {
2836 | bne >1
2837 | add PC, PC, TMP2
2838 } else {
2839 | beq >1
2840 | add PC, PC, TMP2
2841 }
2842 |1:
2843 | ins_next
2844 }
2845 |5: // Either or both types are not numbers.
2846 if (!LJ_DUALNUM) {
2847 | lwz CARG2, 4(RA)
2848 | lwz CARG3, 4(RD)
2849 }
2850 if (LJ_HASFFI) {
2851 | cmpwi cr7, TMP0, LJ_TCDATA
2852 | cmpwi cr5, TMP1, LJ_TCDATA
2853 }
2854 | not TMP3, TMP0
2855 | cmplw TMP0, TMP1
2856 | cmplwi cr1, TMP3, ~LJ_TISPRI // Primitive?
2857 if (LJ_HASFFI) {
2858 | cror 4*cr7+eq, 4*cr7+eq, 4*cr5+eq
2859 }
2860 | cmplwi cr6, TMP3, ~LJ_TISTABUD // Table or userdata?
2861 if (LJ_HASFFI) {
2862 | beq cr7, ->vmeta_equal_cd
2863 }
2864 | cmplw cr5, CARG2, CARG3
2865 | crandc 4*cr0+gt, 4*cr0+eq, 4*cr1+gt // 2: Same type and primitive.
2866 | crorc 4*cr0+lt, 4*cr5+eq, 4*cr0+eq // 1: Same tv or different type.
2867 | crand 4*cr0+eq, 4*cr0+eq, 4*cr5+eq // 0: Same type and same tv.
2868 | mr SAVE0, PC
2869 | cror 4*cr0+eq, 4*cr0+eq, 4*cr0+gt // 0 or 2.
2870 | cror 4*cr0+lt, 4*cr0+lt, 4*cr0+gt // 1 or 2.
2871 if (vk) {
2872 | bne cr0, >6
2873 | add PC, PC, TMP2
2874 |6:
2875 } else {
2876 | beq cr0, >6
2877 | add PC, PC, TMP2
2878 |6:
2879 }
2880 if (LJ_DUALNUM) {
2881 | bge cr0, >2 // Done if 1 or 2.
2882 |1:
2883 | ins_next
2884 |2:
2885 } else {
2886 | blt cr0, <1 // Done if 1 or 2.
2887 }
2888 | blt cr6, <1 // Done if not tab/ud.
2889 |
2890 | // Different tables or userdatas. Need to check __eq metamethod.
2891 | // Field metatable must be at same offset for GCtab and GCudata!
2892 | lwz TAB:TMP2, TAB:CARG2->metatable
2893 | li CARG4, 1-vk // ne = 0 or 1.
2894 | cmplwi TAB:TMP2, 0
2895 | beq <1 // No metatable?
2896 | lbz TMP2, TAB:TMP2->nomm
2897 | andi. TMP2, TMP2, 1<<MM_eq
2898 | bne <1 // Or 'no __eq' flag set?
2899 | mr PC, SAVE0 // Restore old PC.
2900 | b ->vmeta_equal // Handle __eq metamethod.
2901 break;
2902
2903 case BC_ISEQS: case BC_ISNES:
2904 vk = op == BC_ISEQS;
2905 | // RA = src*8, RD = str_const*8 (~), JMP with RD = target
2906 | lwzux TMP0, RA, BASE
2907 | srwi RD, RD, 1
2908 | lwz STR:TMP3, 4(RA)
2909 | lwz TMP2, 0(PC)
2910 | subfic RD, RD, -4
2911 | addi PC, PC, 4
2912 if (LJ_HASFFI) {
2913 | cmpwi TMP0, LJ_TCDATA
2914 }
2915 | lwzx STR:TMP1, KBASE, RD // KBASE-4-str_const*4
2916 | subfic TMP0, TMP0, LJ_TSTR
2917 if (LJ_HASFFI) {
2918 | beq ->vmeta_equal_cd
2919 }
2920 | sub TMP1, STR:TMP1, STR:TMP3
2921 | or TMP0, TMP0, TMP1
2922 | decode_RD4 TMP2, TMP2
2923 | subfic TMP0, TMP0, 0
2924 | addis TMP2, TMP2, -(BCBIAS_J*4 >> 16)
2925 | subfe TMP1, TMP1, TMP1
2926 if (vk) {
2927 | andc TMP2, TMP2, TMP1
2928 } else {
2929 | and TMP2, TMP2, TMP1
2930 }
2931 | add PC, PC, TMP2
2932 | ins_next
2933 break;
2934
2935 case BC_ISEQN: case BC_ISNEN:
2936 vk = op == BC_ISEQN;
2937 | // RA = src*8, RD = num_const*8, JMP with RD = target
2938 if (LJ_DUALNUM) {
2939 | lwzux TMP0, RA, BASE
2940 | addi PC, PC, 4
2941 | lwz CARG2, 4(RA)
2942 | lwzux TMP1, RD, KBASE
2943 | checknum cr0, TMP0
2944 | lwz TMP2, -4(PC)
2945 | checknum cr1, TMP1
2946 | decode_RD4 TMP2, TMP2
2947 | lwz CARG3, 4(RD)
2948 | addis TMP2, TMP2, -(BCBIAS_J*4 >> 16)
2949 if (vk) {
2950 |->BC_ISEQN_Z:
2951 } else {
2952 |->BC_ISNEN_Z:
2953 }
2954 | bne cr0, >7
2955 | bne cr1, >8
2956 | cmpw CARG2, CARG3
2957 |4:
2958 } else {
2959 if (vk) {
2960 |->BC_ISEQN_Z: // Dummy label.
2961 } else {
2962 |->BC_ISNEN_Z: // Dummy label.
2963 }
2964 | lwzx TMP0, BASE, RA
2965 | addi PC, PC, 4
2966 | lfdx f0, BASE, RA
2967 | lwz TMP2, -4(PC)
2968 | lfdx f1, KBASE, RD
2969 | decode_RD4 TMP2, TMP2
2970 | checknum TMP0
2971 | addis TMP2, TMP2, -(BCBIAS_J*4 >> 16)
2972 | bge >3
2973 | fcmpu cr0, f0, f1
2974 }
2975 if (vk) {
2976 | bne >1
2977 | add PC, PC, TMP2
2978 |1:
2979 if (!LJ_HASFFI) {
2980 |3:
2981 }
2982 } else {
2983 | beq >2
2984 |1:
2985 if (!LJ_HASFFI) {
2986 |3:
2987 }
2988 | add PC, PC, TMP2
2989 |2:
2990 }
2991 | ins_next
2992 if (LJ_HASFFI) {
2993 |3:
2994 | cmpwi TMP0, LJ_TCDATA
2995 | beq ->vmeta_equal_cd
2996 | b <1
2997 }
2998 if (LJ_DUALNUM) {
2999 |7: // RA is not an integer.
3000 | bge cr0, <3
3001 | // RA is a number.
3002 | lfd f0, 0(RA)
3003 | blt cr1, >1
3004 | // RA is a number, RD is an integer.
3005 | tonum_i f1, CARG3
3006 | b >2
3007 |
3008 |8: // RA is an integer, RD is a number.
3009 | tonum_i f0, CARG2
3010 |1:
3011 | lfd f1, 0(RD)
3012 |2:
3013 | fcmpu cr0, f0, f1
3014 | b <4
3015 }
3016 break;
3017
3018 case BC_ISEQP: case BC_ISNEP:
3019 vk = op == BC_ISEQP;
3020 | // RA = src*8, RD = primitive_type*8 (~), JMP with RD = target
3021 | lwzx TMP0, BASE, RA
3022 | srwi TMP1, RD, 3
3023 | lwz TMP2, 0(PC)
3024 | not TMP1, TMP1
3025 | addi PC, PC, 4
3026 if (LJ_HASFFI) {
3027 | cmpwi TMP0, LJ_TCDATA
3028 }
3029 | sub TMP0, TMP0, TMP1
3030 if (LJ_HASFFI) {
3031 | beq ->vmeta_equal_cd
3032 }
3033 | decode_RD4 TMP2, TMP2
3034 | addic TMP0, TMP0, -1
3035 | addis TMP2, TMP2, -(BCBIAS_J*4 >> 16)
3036 | subfe TMP1, TMP1, TMP1
3037 if (vk) {
3038 | and TMP2, TMP2, TMP1
3039 } else {
3040 | andc TMP2, TMP2, TMP1
3041 }
3042 | add PC, PC, TMP2
3043 | ins_next
3044 break;
3045
3046 /* -- Unary test and copy ops ------------------------------------------- */
3047
3048 case BC_ISTC: case BC_ISFC: case BC_IST: case BC_ISF:
3049 | // RA = dst*8 or unused, RD = src*8, JMP with RD = target
3050 | lwzx TMP0, BASE, RD
3051 | lwz INS, 0(PC)
3052 | addi PC, PC, 4
3053 if (op == BC_IST || op == BC_ISF) {
3054 | subfic TMP0, TMP0, LJ_TTRUE
3055 | decode_RD4 TMP2, INS
3056 | subfe TMP1, TMP1, TMP1
3057 | addis TMP2, TMP2, -(BCBIAS_J*4 >> 16)
3058 if (op == BC_IST) {
3059 | andc TMP2, TMP2, TMP1
3060 } else {
3061 | and TMP2, TMP2, TMP1
3062 }
3063 | add PC, PC, TMP2
3064 } else {
3065 | li TMP1, LJ_TFALSE
3066 | lfdx f0, BASE, RD
3067 | cmplw TMP0, TMP1
3068 if (op == BC_ISTC) {
3069 | bge >1
3070 } else {
3071 | blt >1
3072 }
3073 | addis PC, PC, -(BCBIAS_J*4 >> 16)
3074 | decode_RD4 TMP2, INS
3075 | stfdx f0, BASE, RA
3076 | add PC, PC, TMP2
3077 |1:
3078 }
3079 | ins_next
3080 break;
3081
3082 /* -- Unary ops --------------------------------------------------------- */
3083
3084 case BC_MOV:
3085 | // RA = dst*8, RD = src*8
3086 | ins_next1
3087 | lfdx f0, BASE, RD
3088 | stfdx f0, BASE, RA
3089 | ins_next2
3090 break;
3091 case BC_NOT:
3092 | // RA = dst*8, RD = src*8
3093 | ins_next1
3094 | lwzx TMP0, BASE, RD
3095 | subfic TMP1, TMP0, LJ_TTRUE
3096 | adde TMP0, TMP0, TMP1
3097 | stwx TMP0, BASE, RA
3098 | ins_next2
3099 break;
3100 case BC_UNM:
3101 | // RA = dst*8, RD = src*8
3102 | lwzux TMP1, RD, BASE
3103 | lwz TMP0, 4(RD)
3104 | checknum TMP1
3105 if (LJ_DUALNUM) {
3106 | bne >5
3107 | nego. TMP0, TMP0
3108 | bso >4
3109 |1:
3110 | ins_next1
3111 | stwux TISNUM, RA, BASE
3112 | stw TMP0, 4(RA)
3113 |3:
3114 | ins_next2
3115 |4: // Potential overflow.
3116 | mcrxr cr0; bley <1 // Ignore unrelated overflow.
3117 | lus TMP1, 0x41e0 // 2^31.
3118 | li TMP0, 0
3119 | b >7
3120 }
3121 |5:
3122 | bge ->vmeta_unm
3123 | xoris TMP1, TMP1, 0x8000
3124 |7:
3125 | ins_next1
3126 | stwux TMP1, RA, BASE
3127 | stw TMP0, 4(RA)
3128 if (LJ_DUALNUM) {
3129 | b <3
3130 } else {
3131 | ins_next2
3132 }
3133 break;
3134 case BC_LEN:
3135 | // RA = dst*8, RD = src*8
3136 | lwzux TMP0, RD, BASE
3137 | lwz CARG1, 4(RD)
3138 | checkstr TMP0; bne >2
3139 | lwz CRET1, STR:CARG1->len
3140 |1:
3141 if (LJ_DUALNUM) {
3142 | ins_next1
3143 | stwux TISNUM, RA, BASE
3144 | stw CRET1, 4(RA)
3145 } else {
3146 | tonum_u f0, CRET1 // Result is a non-negative integer.
3147 | ins_next1
3148 | stfdx f0, BASE, RA
3149 }
3150 | ins_next2
3151 |2:
3152 | checktab TMP0; bne ->vmeta_len
3153#ifdef LUAJIT_ENABLE_LUA52COMPAT
3154 | lwz TAB:TMP2, TAB:CARG1->metatable
3155 | cmplwi TAB:TMP2, 0
3156 | bne >9
3157 |3:
3158#endif
3159 |->BC_LEN_Z:
3160 | bl extern lj_tab_len // (GCtab *t)
3161 | // Returns uint32_t (but less than 2^31).
3162 | b <1
3163#ifdef LUAJIT_ENABLE_LUA52COMPAT
3164 |9:
3165 | lbz TMP0, TAB:TMP2->nomm
3166 | andi. TMP0, TMP0, 1<<MM_len
3167 | bne <3 // 'no __len' flag set: done.
3168 | b ->vmeta_len
3169#endif
3170 break;
3171
3172 /* -- Binary ops -------------------------------------------------------- */
3173
3174 |.macro ins_arithpre
3175 | // RA = dst*8, RB = src1*8, RC = src2*8 | num_const*8
3176 ||vk = ((int)op - BC_ADDVN) / (BC_ADDNV-BC_ADDVN);
3177 ||switch (vk) {
3178 ||case 0:
3179 | lwzx TMP1, BASE, RB
3180 ||if (LJ_DUALNUM) {
3181 | lwzx TMP2, KBASE, RC
3182 ||}
3183 | lfdx f14, BASE, RB
3184 | lfdx f15, KBASE, RC
3185 ||if (LJ_DUALNUM) {
3186 | checknum cr0, TMP1
3187 | checknum cr1, TMP2
3188 | crand 4*cr0+lt, 4*cr0+lt, 4*cr1+lt
3189 | bge ->vmeta_arith_vn
3190 ||} else {
3191 | checknum TMP1; bge ->vmeta_arith_vn
3192 ||}
3193 || break;
3194 ||case 1:
3195 | lwzx TMP1, BASE, RB
3196 ||if (LJ_DUALNUM) {
3197 | lwzx TMP2, KBASE, RC
3198 ||}
3199 | lfdx f15, BASE, RB
3200 | lfdx f14, KBASE, RC
3201 ||if (LJ_DUALNUM) {
3202 | checknum cr0, TMP1
3203 | checknum cr1, TMP2
3204 | crand 4*cr0+lt, 4*cr0+lt, 4*cr1+lt
3205 | bge ->vmeta_arith_nv
3206 ||} else {
3207 | checknum TMP1; bge ->vmeta_arith_nv
3208 ||}
3209 || break;
3210 ||default:
3211 | lwzx TMP1, BASE, RB
3212 | lwzx TMP2, BASE, RC
3213 | lfdx f14, BASE, RB
3214 | lfdx f15, BASE, RC
3215 | checknum cr0, TMP1
3216 | checknum cr1, TMP2
3217 | crand 4*cr0+lt, 4*cr0+lt, 4*cr1+lt
3218 | bge ->vmeta_arith_vv
3219 || break;
3220 ||}
3221 |.endmacro
3222 |
3223 |.macro ins_arithfallback, ins
3224 ||switch (vk) {
3225 ||case 0:
3226 | ins ->vmeta_arith_vn2
3227 || break;
3228 ||case 1:
3229 | ins ->vmeta_arith_nv2
3230 || break;
3231 ||default:
3232 | ins ->vmeta_arith_vv2
3233 || break;
3234 ||}
3235 |.endmacro
3236 |
3237 |.macro intmod, a, b, c
3238 | bl ->vm_modi
3239 |.endmacro
3240 |
3241 |.macro fpmod, a, b, c
3242 |->BC_MODVN_Z:
3243 | fdiv FARG1, b, c
3244 | // NYI: Use internal implementation of floor.
3245 | bl extern floor // floor(b/c)
3246 | fmul a, FARG1, c
3247 | fsub a, b, a // b - floor(b/c)*c
3248 |.endmacro
3249 |
3250 |.macro ins_arithfp, fpins
3251 | ins_arithpre
3252 |.if "fpins" == "fpmod_"
3253 | b ->BC_MODVN_Z // Avoid 3 copies. It's slow anyway.
3254 |.else
3255 | fpins f0, f14, f15
3256 | ins_next1
3257 | stfdx f0, BASE, RA
3258 | ins_next2
3259 |.endif
3260 |.endmacro
3261 |
3262 |.macro ins_arithdn, intins, fpins
3263 | // RA = dst*8, RB = src1*8, RC = src2*8 | num_const*8
3264 ||vk = ((int)op - BC_ADDVN) / (BC_ADDNV-BC_ADDVN);
3265 ||switch (vk) {
3266 ||case 0:
3267 | lwzux TMP1, RB, BASE
3268 | lwzux TMP2, RC, KBASE
3269 | lwz CARG1, 4(RB)
3270 | checknum cr0, TMP1
3271 | lwz CARG2, 4(RC)
3272 || break;
3273 ||case 1:
3274 | lwzux TMP1, RB, BASE
3275 | lwzux TMP2, RC, KBASE
3276 | lwz CARG2, 4(RB)
3277 | checknum cr0, TMP1
3278 | lwz CARG1, 4(RC)
3279 || break;
3280 ||default:
3281 | lwzux TMP1, RB, BASE
3282 | lwzux TMP2, RC, BASE
3283 | lwz CARG1, 4(RB)
3284 | checknum cr0, TMP1
3285 | lwz CARG2, 4(RC)
3286 || break;
3287 ||}
3288 | checknum cr1, TMP2
3289 | bne >5
3290 | bne cr1, >5
3291 | intins CARG1, CARG1, CARG2
3292 | bso >4
3293 |1:
3294 | ins_next1
3295 | stwux TISNUM, RA, BASE
3296 | stw CARG1, 4(RA)
3297 |2:
3298 | ins_next2
3299 |4: // Overflow.
3300 | mcrxr cr0; bley <1 // Ignore unrelated overflow.
3301 | ins_arithfallback b
3302 |5: // FP variant.
3303 ||if (vk == 1) {
3304 | lfd f15, 0(RB)
3305 | crand 4*cr0+lt, 4*cr0+lt, 4*cr1+lt
3306 | lfd f14, 0(RC)
3307 ||} else {
3308 | lfd f14, 0(RB)
3309 | crand 4*cr0+lt, 4*cr0+lt, 4*cr1+lt
3310 | lfd f15, 0(RC)
3311 ||}
3312 | ins_arithfallback bge
3313 |.if "fpins" == "fpmod_"
3314 | b ->BC_MODVN_Z // Avoid 3 copies. It's slow anyway.
3315 |.else
3316 | fpins f0, f14, f15
3317 | ins_next1
3318 | stfdx f0, BASE, RA
3319 | b <2
3320 |.endif
3321 |.endmacro
3322 |
3323 |.macro ins_arith, intins, fpins
3324 ||if (LJ_DUALNUM) {
3325 | ins_arithdn intins, fpins
3326 ||} else {
3327 | ins_arithfp fpins
3328 ||}
3329 |.endmacro
3330
3331 case BC_ADDVN: case BC_ADDNV: case BC_ADDVV:
3332 | ins_arith addo., fadd
3333 break;
3334 case BC_SUBVN: case BC_SUBNV: case BC_SUBVV:
3335 | ins_arith subo., fsub
3336 break;
3337 case BC_MULVN: case BC_MULNV: case BC_MULVV:
3338 | ins_arith mullwo., fmul
3339 break;
3340 case BC_DIVVN: case BC_DIVNV: case BC_DIVVV:
3341 | ins_arithfp fdiv
3342 break;
3343 case BC_MODVN:
3344 | ins_arith intmod, fpmod
3345 break;
3346 case BC_MODNV: case BC_MODVV:
3347 | ins_arith intmod, fpmod_
3348 break;
3349 case BC_POW:
3350 | // NYI: (partial) integer arithmetic.
3351 | lwzx TMP1, BASE, RB
3352 | lfdx FARG1, BASE, RB
3353 | lwzx TMP2, BASE, RC
3354 | lfdx FARG2, BASE, RC
3355 | checknum cr0, TMP1
3356 | checknum cr1, TMP2
3357 | crand 4*cr0+lt, 4*cr0+lt, 4*cr1+lt
3358 | bge ->vmeta_arith_vv
3359 | bl extern pow
3360 | ins_next1
3361 | stfdx FARG1, BASE, RA
3362 | ins_next2
3363 break;
3364
3365 case BC_CAT:
3366 | // RA = dst*8, RB = src_start*8, RC = src_end*8
3367 | sub CARG3, RC, RB
3368 | stw BASE, L->base
3369 | add CARG2, BASE, RC
3370 | mr SAVE0, RB
3371 |->BC_CAT_Z:
3372 | stw PC, SAVE_PC
3373 | mr CARG1, L
3374 | srwi CARG3, CARG3, 3
3375 | bl extern lj_meta_cat // (lua_State *L, TValue *top, int left)
3376 | // Returns NULL (finished) or TValue * (metamethod).
3377 | cmplwi CRET1, 0
3378 | lwz BASE, L->base
3379 | bne ->vmeta_binop
3380 | ins_next1
3381 | lfdx f0, BASE, SAVE0 // Copy result from RB to RA.
3382 | stfdx f0, BASE, RA
3383 | ins_next2
3384 break;
3385
3386 /* -- Constant ops ------------------------------------------------------ */
3387
3388 case BC_KSTR:
3389 | // RA = dst*8, RD = str_const*8 (~)
3390 | srwi TMP1, RD, 1
3391 | subfic TMP1, TMP1, -4
3392 | ins_next1
3393 | lwzx TMP0, KBASE, TMP1 // KBASE-4-str_const*4
3394 | li TMP2, LJ_TSTR
3395 | stwux TMP2, RA, BASE
3396 | stw TMP0, 4(RA)
3397 | ins_next2
3398 break;
3399 case BC_KCDATA:
3400#if LJ_HASFFI
3401 | // RA = dst*8, RD = cdata_const*8 (~)
3402 | srwi TMP1, RD, 1
3403 | subfic TMP1, TMP1, -4
3404 | ins_next1
3405 | lwzx TMP0, KBASE, TMP1 // KBASE-4-cdata_const*4
3406 | li TMP2, LJ_TCDATA
3407 | stwux TMP2, RA, BASE
3408 | stw TMP0, 4(RA)
3409 | ins_next2
3410#endif
3411 break;
3412 case BC_KSHORT:
3413 | // RA = dst*8, RD = int16_literal*8
3414 if (LJ_DUALNUM) {
3415 | slwi RD, RD, 13
3416 | srawi RD, RD, 16
3417 | ins_next1
3418 | stwux TISNUM, RA, BASE
3419 | stw RD, 4(RA)
3420 | ins_next2
3421 } else {
3422 | // The soft-float approach is faster.
3423 | slwi RD, RD, 13
3424 | srawi TMP1, RD, 31
3425 | xor TMP2, TMP1, RD
3426 | sub TMP2, TMP2, TMP1 // TMP2 = abs(x)
3427 | cntlzw TMP3, TMP2
3428 | subfic TMP1, TMP3, 0x40d // TMP1 = exponent-1
3429 | slw TMP2, TMP2, TMP3 // TMP2 = left aligned mantissa
3430 | subfic TMP3, RD, 0
3431 | slwi TMP1, TMP1, 20
3432 | rlwimi RD, TMP2, 21, 1, 31 // hi = sign(x) | (mantissa>>11)
3433 | subfe TMP0, TMP0, TMP0
3434 | add RD, RD, TMP1 // hi = hi + exponent-1
3435 | and RD, RD, TMP0 // hi = x == 0 ? 0 : hi
3436 | ins_next1
3437 | stwux RD, RA, BASE
3438 | stw ZERO, 4(RA)
3439 | ins_next2
3440 }
3441 break;
3442 case BC_KNUM:
3443 | // RA = dst*8, RD = num_const*8
3444 | ins_next1
3445 | lfdx f0, KBASE, RD
3446 | stfdx f0, BASE, RA
3447 | ins_next2
3448 break;
3449 case BC_KPRI:
3450 | // RA = dst*8, RD = primitive_type*8 (~)
3451 | srwi TMP1, RD, 3
3452 | not TMP0, TMP1
3453 | ins_next1
3454 | stwx TMP0, BASE, RA
3455 | ins_next2
3456 break;
3457 case BC_KNIL:
3458 | // RA = base*8, RD = end*8
3459 | stwx TISNIL, BASE, RA
3460 | addi RA, RA, 8
3461 |1:
3462 | stwx TISNIL, BASE, RA
3463 | cmpw RA, RD
3464 | addi RA, RA, 8
3465 | blt <1
3466 | ins_next_
3467 break;
3468
3469 /* -- Upvalue and function ops ------------------------------------------ */
3470
3471 case BC_UGET:
3472 | // RA = dst*8, RD = uvnum*8
3473 | lwz LFUNC:RB, FRAME_FUNC(BASE)
3474 | srwi RD, RD, 1
3475 | addi RD, RD, offsetof(GCfuncL, uvptr)
3476 | lwzx UPVAL:RB, LFUNC:RB, RD
3477 | ins_next1
3478 | lwz TMP1, UPVAL:RB->v
3479 | lfd f0, 0(TMP1)
3480 | stfdx f0, BASE, RA
3481 | ins_next2
3482 break;
3483 case BC_USETV:
3484 | // RA = uvnum*8, RD = src*8
3485 | lwz LFUNC:RB, FRAME_FUNC(BASE)
3486 | srwi RA, RA, 1
3487 | addi RA, RA, offsetof(GCfuncL, uvptr)
3488 | lfdux f0, RD, BASE
3489 | lwzx UPVAL:RB, LFUNC:RB, RA
3490 | lbz TMP3, UPVAL:RB->marked
3491 | lwz CARG2, UPVAL:RB->v
3492 | andi. TMP3, TMP3, LJ_GC_BLACK // isblack(uv)
3493 | lbz TMP0, UPVAL:RB->closed
3494 | lwz TMP2, 0(RD)
3495 | stfd f0, 0(CARG2)
3496 | cmplwi cr1, TMP0, 0
3497 | lwz TMP1, 4(RD)
3498 | cror 4*cr0+eq, 4*cr0+eq, 4*cr1+eq
3499 | subi TMP2, TMP2, (LJ_TISNUM+1)
3500 | bne >2 // Upvalue is closed and black?
3501 |1:
3502 | ins_next
3503 |
3504 |2: // Check if new value is collectable.
3505 | cmplwi TMP2, LJ_TISGCV - (LJ_TISNUM+1)
3506 | bge <1 // tvisgcv(v)
3507 | lbz TMP3, GCOBJ:TMP1->gch.marked
3508 | andi. TMP3, TMP3, LJ_GC_WHITES // iswhite(v)
3509 | la CARG1, GG_DISP2G(DISPATCH)
3510 | // Crossed a write barrier. Move the barrier forward.
3511 | beq <1
3512 | bl extern lj_gc_barrieruv // (global_State *g, TValue *tv)
3513 | b <1
3514 break;
3515 case BC_USETS:
3516 | // RA = uvnum*8, RD = str_const*8 (~)
3517 | lwz LFUNC:RB, FRAME_FUNC(BASE)
3518 | srwi TMP1, RD, 1
3519 | srwi RA, RA, 1
3520 | subfic TMP1, TMP1, -4
3521 | addi RA, RA, offsetof(GCfuncL, uvptr)
3522 | lwzx STR:TMP1, KBASE, TMP1 // KBASE-4-str_const*4
3523 | lwzx UPVAL:RB, LFUNC:RB, RA
3524 | lbz TMP3, UPVAL:RB->marked
3525 | lwz CARG2, UPVAL:RB->v
3526 | andi. TMP3, TMP3, LJ_GC_BLACK // isblack(uv)
3527 | lbz TMP3, STR:TMP1->marked
3528 | lbz TMP2, UPVAL:RB->closed
3529 | li TMP0, LJ_TSTR
3530 | stw STR:TMP1, 4(CARG2)
3531 | stw TMP0, 0(CARG2)
3532 | bne >2
3533 |1:
3534 | ins_next
3535 |
3536 |2: // Check if string is white and ensure upvalue is closed.
3537 | andi. TMP3, TMP3, LJ_GC_WHITES // iswhite(str)
3538 | cmplwi cr1, TMP2, 0
3539 | cror 4*cr0+eq, 4*cr0+eq, 4*cr1+eq
3540 | la CARG1, GG_DISP2G(DISPATCH)
3541 | // Crossed a write barrier. Move the barrier forward.
3542 | beq <1
3543 | bl extern lj_gc_barrieruv // (global_State *g, TValue *tv)
3544 | b <1
3545 break;
3546 case BC_USETN:
3547 | // RA = uvnum*8, RD = num_const*8
3548 | lwz LFUNC:RB, FRAME_FUNC(BASE)
3549 | srwi RA, RA, 1
3550 | addi RA, RA, offsetof(GCfuncL, uvptr)
3551 | lfdx f0, KBASE, RD
3552 | lwzx UPVAL:RB, LFUNC:RB, RA
3553 | ins_next1
3554 | lwz TMP1, UPVAL:RB->v
3555 | stfd f0, 0(TMP1)
3556 | ins_next2
3557 break;
3558 case BC_USETP:
3559 | // RA = uvnum*8, RD = primitive_type*8 (~)
3560 | lwz LFUNC:RB, FRAME_FUNC(BASE)
3561 | srwi RA, RA, 1
3562 | srwi TMP0, RD, 3
3563 | addi RA, RA, offsetof(GCfuncL, uvptr)
3564 | not TMP0, TMP0
3565 | lwzx UPVAL:RB, LFUNC:RB, RA
3566 | ins_next1
3567 | lwz TMP1, UPVAL:RB->v
3568 | stw TMP0, 0(TMP1)
3569 | ins_next2
3570 break;
3571
3572 case BC_UCLO:
3573 | // RA = level*8, RD = target
3574 | lwz TMP1, L->openupval
3575 | branch_RD // Do this first since RD is not saved.
3576 | stw BASE, L->base
3577 | cmplwi TMP1, 0
3578 | mr CARG1, L
3579 | beq >1
3580 | add CARG2, BASE, RA
3581 | bl extern lj_func_closeuv // (lua_State *L, TValue *level)
3582 | lwz BASE, L->base
3583 |1:
3584 | ins_next
3585 break;
3586
3587 case BC_FNEW:
3588 | // RA = dst*8, RD = proto_const*8 (~) (holding function prototype)
3589 | srwi TMP1, RD, 1
3590 | stw BASE, L->base
3591 | subfic TMP1, TMP1, -4
3592 | stw PC, SAVE_PC
3593 | lwzx CARG2, KBASE, TMP1 // KBASE-4-tab_const*4
3594 | mr CARG1, L
3595 | lwz CARG3, FRAME_FUNC(BASE)
3596 | // (lua_State *L, GCproto *pt, GCfuncL *parent)
3597 | bl extern lj_func_newL_gc
3598 | // Returns GCfuncL *.
3599 | lwz BASE, L->base
3600 | li TMP0, LJ_TFUNC
3601 | stwux TMP0, RA, BASE
3602 | stw LFUNC:CRET1, 4(RA)
3603 | ins_next
3604 break;
3605
3606 /* -- Table ops --------------------------------------------------------- */
3607
3608 case BC_TNEW:
3609 case BC_TDUP:
3610 | // RA = dst*8, RD = (hbits|asize)*8 | tab_const*8 (~)
3611 | lwz TMP0, DISPATCH_GL(gc.total)(DISPATCH)
3612 | mr CARG1, L
3613 | lwz TMP1, DISPATCH_GL(gc.threshold)(DISPATCH)
3614 | stw BASE, L->base
3615 | cmplw TMP0, TMP1
3616 | stw PC, SAVE_PC
3617 | bge >5
3618 |1:
3619 if (op == BC_TNEW) {
3620 | rlwinm CARG2, RD, 29, 21, 31
3621 | rlwinm CARG3, RD, 18, 27, 31
3622 | cmpwi CARG2, 0x7ff; beq >3
3623 |2:
3624 | bl extern lj_tab_new // (lua_State *L, int32_t asize, uint32_t hbits)
3625 | // Returns Table *.
3626 } else {
3627 | srwi TMP1, RD, 1
3628 | subfic TMP1, TMP1, -4
3629 | lwzx CARG2, KBASE, TMP1 // KBASE-4-tab_const*4
3630 | bl extern lj_tab_dup // (lua_State *L, Table *kt)
3631 | // Returns Table *.
3632 }
3633 | lwz BASE, L->base
3634 | li TMP0, LJ_TTAB
3635 | stwux TMP0, RA, BASE
3636 | stw TAB:CRET1, 4(RA)
3637 | ins_next
3638 if (op == BC_TNEW) {
3639 |3:
3640 | li CARG2, 0x801
3641 | b <2
3642 }
3643 |5:
3644 | mr SAVE0, RD
3645 | bl extern lj_gc_step_fixtop // (lua_State *L)
3646 | mr RD, SAVE0
3647 | mr CARG1, L
3648 | b <1
3649 break;
3650
3651 case BC_GGET:
3652 | // RA = dst*8, RD = str_const*8 (~)
3653 case BC_GSET:
3654 | // RA = src*8, RD = str_const*8 (~)
3655 | lwz LFUNC:TMP2, FRAME_FUNC(BASE)
3656 | srwi TMP1, RD, 1
3657 | lwz TAB:RB, LFUNC:TMP2->env
3658 | subfic TMP1, TMP1, -4
3659 | lwzx STR:RC, KBASE, TMP1 // KBASE-4-str_const*4
3660 if (op == BC_GGET) {
3661 | b ->BC_TGETS_Z
3662 } else {
3663 | b ->BC_TSETS_Z
3664 }
3665 break;
3666
3667 case BC_TGETV:
3668 | // RA = dst*8, RB = table*8, RC = key*8
3669 | lwzux CARG1, RB, BASE
3670 | lwzux CARG2, RC, BASE
3671 | lwz TAB:RB, 4(RB)
3672 if (LJ_DUALNUM) {
3673 | lwz RC, 4(RC)
3674 } else {
3675 | lfd f0, 0(RC)
3676 }
3677 | checktab CARG1
3678 | checknum cr1, CARG2
3679 | bne ->vmeta_tgetv
3680 if (LJ_DUALNUM) {
3681 | lwz TMP0, TAB:RB->asize
3682 | bne cr1, >5
3683 | lwz TMP1, TAB:RB->array
3684 | cmplw TMP0, RC
3685 | slwi TMP2, RC, 3
3686 } else {
3687 | bge cr1, >5
3688 | // Convert number key to integer, check for integerness and range.
3689 | fctiwz f1, f0
3690 | fadd f2, f0, TOBIT
3691 | stfd f1, TMPD
3692 | lwz TMP0, TAB:RB->asize
3693 | fsub f2, f2, TOBIT
3694 | lwz TMP2, TMPD_LO
3695 | lwz TMP1, TAB:RB->array
3696 | fcmpu cr1, f0, f2
3697 | cmplw cr0, TMP0, TMP2
3698 | crand 4*cr0+gt, 4*cr0+gt, 4*cr1+eq
3699 | slwi TMP2, TMP2, 3
3700 }
3701 | ble ->vmeta_tgetv // Integer key and in array part?
3702 | lwzx TMP0, TMP1, TMP2
3703 | lfdx f14, TMP1, TMP2
3704 | checknil TMP0; beq >2
3705 |1:
3706 | ins_next1
3707 | stfdx f14, BASE, RA
3708 | ins_next2
3709 |
3710 |2: // Check for __index if table value is nil.
3711 | lwz TAB:TMP2, TAB:RB->metatable
3712 | cmplwi TAB:TMP2, 0
3713 | beq <1 // No metatable: done.
3714 | lbz TMP0, TAB:TMP2->nomm
3715 | andi. TMP0, TMP0, 1<<MM_index
3716 | bne <1 // 'no __index' flag set: done.
3717 | b ->vmeta_tgetv
3718 |
3719 |5:
3720 | checkstr CARG2; bne ->vmeta_tgetv
3721 if (!LJ_DUALNUM) {
3722 | lwz STR:RC, 4(RC)
3723 }
3724 | b ->BC_TGETS_Z // String key?
3725 break;
3726 case BC_TGETS:
3727 | // RA = dst*8, RB = table*8, RC = str_const*8 (~)
3728 | lwzux CARG1, RB, BASE
3729 | srwi TMP1, RC, 1
3730 | lwz TAB:RB, 4(RB)
3731 | subfic TMP1, TMP1, -4
3732 | checktab CARG1
3733 | lwzx STR:RC, KBASE, TMP1 // KBASE-4-str_const*4
3734 | bne ->vmeta_tgets1
3735 |->BC_TGETS_Z:
3736 | // TAB:RB = GCtab *, STR:RC = GCstr *, RA = dst*8
3737 | lwz TMP0, TAB:RB->hmask
3738 | lwz TMP1, STR:RC->hash
3739 | lwz NODE:TMP2, TAB:RB->node
3740 | and TMP1, TMP1, TMP0 // idx = str->hash & tab->hmask
3741 | slwi TMP0, TMP1, 5
3742 | slwi TMP1, TMP1, 3
3743 | sub TMP1, TMP0, TMP1
3744 | add NODE:TMP2, NODE:TMP2, TMP1 // node = tab->node + (idx*32-idx*8)
3745 |1:
3746 | lwz CARG1, NODE:TMP2->key
3747 | lwz TMP0, 4+offsetof(Node, key)(NODE:TMP2)
3748 | lwz CARG2, NODE:TMP2->val
3749 | lwz TMP1, 4+offsetof(Node, val)(NODE:TMP2)
3750 | checkstr CARG1; bne >4
3751 | cmpw TMP0, STR:RC; bne >4
3752 | checknil CARG2; beq >5 // Key found, but nil value?
3753 |3:
3754 | stwux CARG2, RA, BASE
3755 | stw TMP1, 4(RA)
3756 | ins_next
3757 |
3758 |4: // Follow hash chain.
3759 | lwz NODE:TMP2, NODE:TMP2->next
3760 | cmplwi NODE:TMP2, 0
3761 | bne <1
3762 | // End of hash chain: key not found, nil result.
3763 | li CARG2, LJ_TNIL
3764 |
3765 |5: // Check for __index if table value is nil.
3766 | lwz TAB:TMP2, TAB:RB->metatable
3767 | cmplwi TAB:TMP2, 0
3768 | beq <3 // No metatable: done.
3769 | lbz TMP0, TAB:TMP2->nomm
3770 | andi. TMP0, TMP0, 1<<MM_index
3771 | bne <3 // 'no __index' flag set: done.
3772 | b ->vmeta_tgets
3773 break;
3774 case BC_TGETB:
3775 | // RA = dst*8, RB = table*8, RC = index*8
3776 | lwzux CARG1, RB, BASE
3777 | srwi TMP0, RC, 3
3778 | lwz TAB:RB, 4(RB)
3779 | checktab CARG1; bne ->vmeta_tgetb
3780 | lwz TMP1, TAB:RB->asize
3781 | lwz TMP2, TAB:RB->array
3782 | cmplw TMP0, TMP1; bge ->vmeta_tgetb
3783 | lwzx TMP1, TMP2, RC
3784 | lfdx f0, TMP2, RC
3785 | checknil TMP1; beq >5
3786 |1:
3787 | ins_next1
3788 | stfdx f0, BASE, RA
3789 | ins_next2
3790 |
3791 |5: // Check for __index if table value is nil.
3792 | lwz TAB:TMP2, TAB:RB->metatable
3793 | cmplwi TAB:TMP2, 0
3794 | beq <1 // No metatable: done.
3795 | lbz TMP2, TAB:TMP2->nomm
3796 | andi. TMP2, TMP2, 1<<MM_index
3797 | bne <1 // 'no __index' flag set: done.
3798 | b ->vmeta_tgetb // Caveat: preserve TMP0!
3799 break;
3800
3801 case BC_TSETV:
3802 | // RA = src*8, RB = table*8, RC = key*8
3803 | lwzux CARG1, RB, BASE
3804 | lwzux CARG2, RC, BASE
3805 | lwz TAB:RB, 4(RB)
3806 if (LJ_DUALNUM) {
3807 | lwz RC, 4(RC)
3808 } else {
3809 | lfd f0, 0(RC)
3810 }
3811 | checktab CARG1
3812 | checknum cr1, CARG2
3813 | bne ->vmeta_tsetv
3814 if (LJ_DUALNUM) {
3815 | lwz TMP0, TAB:RB->asize
3816 | bne cr1, >5
3817 | lwz TMP1, TAB:RB->array
3818 | cmplw TMP0, RC
3819 | slwi TMP0, RC, 3
3820 } else {
3821 | bge cr1, >5
3822 | // Convert number key to integer, check for integerness and range.
3823 | fctiwz f1, f0
3824 | fadd f2, f0, TOBIT
3825 | stfd f1, TMPD
3826 | lwz TMP0, TAB:RB->asize
3827 | fsub f2, f2, TOBIT
3828 | lwz TMP2, TMPD_LO
3829 | lwz TMP1, TAB:RB->array
3830 | fcmpu cr1, f0, f2
3831 | cmplw cr0, TMP0, TMP2
3832 | crand 4*cr0+gt, 4*cr0+gt, 4*cr1+eq
3833 | slwi TMP0, TMP2, 3
3834 }
3835 | ble ->vmeta_tsetv // Integer key and in array part?
3836 | lwzx TMP2, TMP1, TMP0
3837 | lbz TMP3, TAB:RB->marked
3838 | lfdx f14, BASE, RA
3839 | checknil TMP2; beq >3
3840 |1:
3841 | andi. TMP2, TMP3, LJ_GC_BLACK // isblack(table)
3842 | stfdx f14, TMP1, TMP0
3843 | bne >7
3844 |2:
3845 | ins_next
3846 |
3847 |3: // Check for __newindex if previous value is nil.
3848 | lwz TAB:TMP2, TAB:RB->metatable
3849 | cmplwi TAB:TMP2, 0
3850 | beq <1 // No metatable: done.
3851 | lbz TMP2, TAB:TMP2->nomm
3852 | andi. TMP2, TMP2, 1<<MM_newindex
3853 | bne <1 // 'no __newindex' flag set: done.
3854 | b ->vmeta_tsetv
3855 |
3856 |5:
3857 | checkstr CARG2; bne ->vmeta_tsetv
3858 if (!LJ_DUALNUM) {
3859 | lwz STR:RC, 4(RC)
3860 }
3861 | b ->BC_TSETS_Z // String key?
3862 |
3863 |7: // Possible table write barrier for the value. Skip valiswhite check.
3864 | barrierback TAB:RB, TMP3, TMP0
3865 | b <2
3866 break;
3867 case BC_TSETS:
3868 | // RA = src*8, RB = table*8, RC = str_const*8 (~)
3869 | lwzux CARG1, RB, BASE
3870 | srwi TMP1, RC, 1
3871 | lwz TAB:RB, 4(RB)
3872 | subfic TMP1, TMP1, -4
3873 | checktab CARG1
3874 | lwzx STR:RC, KBASE, TMP1 // KBASE-4-str_const*4
3875 | bne ->vmeta_tsets1
3876 |->BC_TSETS_Z:
3877 | // TAB:RB = GCtab *, STR:RC = GCstr *, RA = src*8
3878 | lwz TMP0, TAB:RB->hmask
3879 | lwz TMP1, STR:RC->hash
3880 | lwz NODE:TMP2, TAB:RB->node
3881 | stb ZERO, TAB:RB->nomm // Clear metamethod cache.
3882 | and TMP1, TMP1, TMP0 // idx = str->hash & tab->hmask
3883 | lfdx f14, BASE, RA
3884 | slwi TMP0, TMP1, 5
3885 | slwi TMP1, TMP1, 3
3886 | sub TMP1, TMP0, TMP1
3887 | lbz TMP3, TAB:RB->marked
3888 | add NODE:TMP2, NODE:TMP2, TMP1 // node = tab->node + (idx*32-idx*8)
3889 |1:
3890 | lwz CARG1, NODE:TMP2->key
3891 | lwz TMP0, 4+offsetof(Node, key)(NODE:TMP2)
3892 | lwz CARG2, NODE:TMP2->val
3893 | lwz NODE:TMP1, NODE:TMP2->next
3894 | checkstr CARG1; bne >5
3895 | cmpw TMP0, STR:RC; bne >5
3896 | checknil CARG2; beq >4 // Key found, but nil value?
3897 |2:
3898 | andi. TMP0, TMP3, LJ_GC_BLACK // isblack(table)
3899 | stfd f14, NODE:TMP2->val
3900 | bne >7
3901 |3:
3902 | ins_next
3903 |
3904 |4: // Check for __newindex if previous value is nil.
3905 | lwz TAB:TMP1, TAB:RB->metatable
3906 | cmplwi TAB:TMP1, 0
3907 | beq <2 // No metatable: done.
3908 | lbz TMP0, TAB:TMP1->nomm
3909 | andi. TMP0, TMP0, 1<<MM_newindex
3910 | bne <2 // 'no __newindex' flag set: done.
3911 | b ->vmeta_tsets
3912 |
3913 |5: // Follow hash chain.
3914 | cmplwi NODE:TMP1, 0
3915 | mr NODE:TMP2, NODE:TMP1
3916 | bne <1
3917 | // End of hash chain: key not found, add a new one.
3918 |
3919 | // But check for __newindex first.
3920 | lwz TAB:TMP1, TAB:RB->metatable
3921 | la CARG3, DISPATCH_GL(tmptv)(DISPATCH)
3922 | stw PC, SAVE_PC
3923 | mr CARG1, L
3924 | cmplwi TAB:TMP1, 0
3925 | stw BASE, L->base
3926 | beq >6 // No metatable: continue.
3927 | lbz TMP0, TAB:TMP1->nomm
3928 | andi. TMP0, TMP0, 1<<MM_newindex
3929 | beq ->vmeta_tsets // 'no __newindex' flag NOT set: check.
3930 |6:
3931 | li TMP0, LJ_TSTR
3932 | stw STR:RC, 4(CARG3)
3933 | mr CARG2, TAB:RB
3934 | stw TMP0, 0(CARG3)
3935 | bl extern lj_tab_newkey // (lua_State *L, GCtab *t, TValue *k)
3936 | // Returns TValue *.
3937 | lwz BASE, L->base
3938 | stfd f14, 0(CRET1)
3939 | b <3 // No 2nd write barrier needed.
3940 |
3941 |7: // Possible table write barrier for the value. Skip valiswhite check.
3942 | barrierback TAB:RB, TMP3, TMP0
3943 | b <3
3944 break;
3945 case BC_TSETB:
3946 | // RA = src*8, RB = table*8, RC = index*8
3947 | lwzux CARG1, RB, BASE
3948 | srwi TMP0, RC, 3
3949 | lwz TAB:RB, 4(RB)
3950 | checktab CARG1; bne ->vmeta_tsetb
3951 | lwz TMP1, TAB:RB->asize
3952 | lwz TMP2, TAB:RB->array
3953 | lbz TMP3, TAB:RB->marked
3954 | cmplw TMP0, TMP1
3955 | lfdx f14, BASE, RA
3956 | bge ->vmeta_tsetb
3957 | lwzx TMP1, TMP2, RC
3958 | checknil TMP1; beq >5
3959 |1:
3960 | andi. TMP0, TMP3, LJ_GC_BLACK // isblack(table)
3961 | stfdx f14, TMP2, RC
3962 | bne >7
3963 |2:
3964 | ins_next
3965 |
3966 |5: // Check for __newindex if previous value is nil.
3967 | lwz TAB:TMP1, TAB:RB->metatable
3968 | cmplwi TAB:TMP1, 0
3969 | beq <1 // No metatable: done.
3970 | lbz TMP1, TAB:TMP1->nomm
3971 | andi. TMP1, TMP1, 1<<MM_newindex
3972 | bne <1 // 'no __newindex' flag set: done.
3973 | b ->vmeta_tsetb // Caveat: preserve TMP0!
3974 |
3975 |7: // Possible table write barrier for the value. Skip valiswhite check.
3976 | barrierback TAB:RB, TMP3, TMP0
3977 | b <2
3978 break;
3979
3980 case BC_TSETM:
3981 | // RA = base*8 (table at base-1), RD = num_const*8 (start index)
3982 | add RA, BASE, RA
3983 |1:
3984 | add TMP3, KBASE, RD
3985 | lwz TAB:CARG2, -4(RA) // Guaranteed to be a table.
3986 | addic. TMP0, MULTRES, -8
3987 | lwz TMP3, 4(TMP3) // Integer constant is in lo-word.
3988 | srwi CARG3, TMP0, 3
3989 | beq >4 // Nothing to copy?
3990 | add CARG3, CARG3, TMP3
3991 | lwz TMP2, TAB:CARG2->asize
3992 | slwi TMP1, TMP3, 3
3993 | lbz TMP3, TAB:CARG2->marked
3994 | cmplw CARG3, TMP2
3995 | add TMP2, RA, TMP0
3996 | lwz TMP0, TAB:CARG2->array
3997 | bgt >5
3998 | add TMP1, TMP1, TMP0
3999 | andi. TMP0, TMP3, LJ_GC_BLACK // isblack(table)
4000 |3: // Copy result slots to table.
4001 | lfd f0, 0(RA)
4002 | addi RA, RA, 8
4003 | cmpw cr1, RA, TMP2
4004 | stfd f0, 0(TMP1)
4005 | addi TMP1, TMP1, 8
4006 | blt cr1, <3
4007 | bne >7
4008 |4:
4009 | ins_next
4010 |
4011 |5: // Need to resize array part.
4012 | stw BASE, L->base
4013 | mr CARG1, L
4014 | stw PC, SAVE_PC
4015 | mr SAVE0, RD
4016 | bl extern lj_tab_reasize // (lua_State *L, GCtab *t, int nasize)
4017 | // Must not reallocate the stack.
4018 | mr RD, SAVE0
4019 | b <1
4020 |
4021 |7: // Possible table write barrier for any value. Skip valiswhite check.
4022 | barrierback TAB:CARG2, TMP3, TMP0
4023 | b <4
4024 break;
4025
4026 /* -- Calls and vararg handling ----------------------------------------- */
4027
4028 case BC_CALLM:
4029 | // RA = base*8, (RB = (nresults+1)*8,) RC = extra_nargs*8
4030 | add NARGS8:RC, NARGS8:RC, MULTRES
4031 | // Fall through. Assumes BC_CALL follows.
4032 break;
4033 case BC_CALL:
4034 | // RA = base*8, (RB = (nresults+1)*8,) RC = (nargs+1)*8
4035 | mr TMP2, BASE
4036 | lwzux TMP0, BASE, RA
4037 | lwz LFUNC:RB, 4(BASE)
4038 | subi NARGS8:RC, NARGS8:RC, 8
4039 | addi BASE, BASE, 8
4040 | checkfunc TMP0; bne ->vmeta_call
4041 | ins_call
4042 break;
4043
4044 case BC_CALLMT:
4045 | // RA = base*8, (RB = 0,) RC = extra_nargs*8
4046 | add NARGS8:RC, NARGS8:RC, MULTRES
4047 | // Fall through. Assumes BC_CALLT follows.
4048 break;
4049 case BC_CALLT:
4050 | // RA = base*8, (RB = 0,) RC = (nargs+1)*8
4051 | lwzux TMP0, RA, BASE
4052 | lwz LFUNC:RB, 4(RA)
4053 | subi NARGS8:RC, NARGS8:RC, 8
4054 | lwz TMP1, FRAME_PC(BASE)
4055 | checkfunc TMP0
4056 | addi RA, RA, 8
4057 | bne ->vmeta_callt
4058 |->BC_CALLT_Z:
4059 | andi. TMP0, TMP1, FRAME_TYPE // Caveat: preserve cr0 until the crand.
4060 | lbz TMP3, LFUNC:RB->ffid
4061 | xori TMP2, TMP1, FRAME_VARG
4062 | cmplwi cr1, NARGS8:RC, 0
4063 | bne >7
4064 |1:
4065 | stw LFUNC:RB, FRAME_FUNC(BASE) // Copy function down, but keep PC.
4066 | li TMP2, 0
4067 | cmplwi cr7, TMP3, 1 // (> FF_C) Calling a fast function?
4068 | beq cr1, >3
4069 |2:
4070 | addi TMP3, TMP2, 8
4071 | lfdx f0, RA, TMP2
4072 | cmplw cr1, TMP3, NARGS8:RC
4073 | stfdx f0, BASE, TMP2
4074 | mr TMP2, TMP3
4075 | bne cr1, <2
4076 |3:
4077 | crand 4*cr0+eq, 4*cr0+eq, 4*cr7+gt
4078 | beq >5
4079 |4:
4080 | ins_callt
4081 |
4082 |5: // Tailcall to a fast function with a Lua frame below.
4083 | lwz INS, -4(TMP1)
4084 | decode_RA8 RA, INS
4085 | sub TMP1, BASE, RA
4086 | lwz LFUNC:TMP1, FRAME_FUNC-8(TMP1)
4087 | lwz TMP1, LFUNC:TMP1->pc
4088 | lwz KBASE, PC2PROTO(k)(TMP1) // Need to prepare KBASE.
4089 | b <4
4090 |
4091 |7: // Tailcall from a vararg function.
4092 | andi. TMP0, TMP2, FRAME_TYPEP
4093 | bne <1 // Vararg frame below?
4094 | sub BASE, BASE, TMP2 // Relocate BASE down.
4095 | lwz TMP1, FRAME_PC(BASE)
4096 | andi. TMP0, TMP1, FRAME_TYPE
4097 | b <1
4098 break;
4099
4100 case BC_ITERC:
4101 | // RA = base*8, (RB = (nresults+1)*8, RC = (nargs+1)*8 ((2+1)*8))
4102 | mr TMP2, BASE
4103 | add BASE, BASE, RA
4104 | lwz TMP1, -24(BASE)
4105 | lwz LFUNC:RB, -20(BASE)
4106 | lfd f1, -8(BASE)
4107 | lfd f0, -16(BASE)
4108 | stw TMP1, 0(BASE) // Copy callable.
4109 | stw LFUNC:RB, 4(BASE)
4110 | checkfunc TMP1
4111 | stfd f1, 16(BASE) // Copy control var.
4112 | li NARGS8:RC, 16 // Iterators get 2 arguments.
4113 | stfdu f0, 8(BASE) // Copy state.
4114 | bne ->vmeta_call
4115 | ins_call
4116 break;
4117
4118 case BC_ITERN:
4119 | // RA = base*8, (RB = (nresults+1)*8, RC = (nargs+1)*8 (2+1)*8)
4120#if LJ_HASJIT
4121 | // NYI: add hotloop, record BC_ITERN.
4122#endif
4123 | add RA, BASE, RA
4124 | lwz TAB:RB, -12(RA)
4125 | lwz RC, -4(RA) // Get index from control var.
4126 | lwz TMP0, TAB:RB->asize
4127 | lwz TMP1, TAB:RB->array
4128 | addi PC, PC, 4
4129 |1: // Traverse array part.
4130 | cmplw RC, TMP0
4131 | slwi TMP3, RC, 3
4132 | bge >5 // Index points after array part?
4133 | lwzx TMP2, TMP1, TMP3
4134 | lfdx f0, TMP1, TMP3
4135 | checknil TMP2
4136 | lwz INS, -4(PC)
4137 | beq >4
4138 if (LJ_DUALNUM) {
4139 | stw RC, 4(RA)
4140 | stw TISNUM, 0(RA)
4141 } else {
4142 | tonum_u f1, RC
4143 }
4144 | addi RC, RC, 1
4145 | addis TMP3, PC, -(BCBIAS_J*4 >> 16)
4146 | stfd f0, 8(RA)
4147 | decode_RD4 TMP1, INS
4148 | stw RC, -4(RA) // Update control var.
4149 | add PC, TMP1, TMP3
4150 if (!LJ_DUALNUM) {
4151 | stfd f1, 0(RA)
4152 }
4153 |3:
4154 | ins_next
4155 |
4156 |4: // Skip holes in array part.
4157 | addi RC, RC, 1
4158 | b <1
4159 |
4160 |5: // Traverse hash part.
4161 | lwz TMP1, TAB:RB->hmask
4162 | sub RC, RC, TMP0
4163 | lwz TMP2, TAB:RB->node
4164 |6:
4165 | cmplw RC, TMP1 // End of iteration? Branch to ITERL+1.
4166 | slwi TMP3, RC, 5
4167 | bgty <3
4168 | slwi RB, RC, 3
4169 | sub TMP3, TMP3, RB
4170 | lwzx RB, TMP2, TMP3
4171 | lfdx f0, TMP2, TMP3
4172 | add NODE:TMP3, TMP2, TMP3
4173 | checknil RB
4174 | lwz INS, -4(PC)
4175 | beq >7
4176 | lfd f1, NODE:TMP3->key
4177 | addis TMP2, PC, -(BCBIAS_J*4 >> 16)
4178 | stfd f0, 8(RA)
4179 | add RC, RC, TMP0
4180 | decode_RD4 TMP1, INS
4181 | stfd f1, 0(RA)
4182 | addi RC, RC, 1
4183 | add PC, TMP1, TMP2
4184 | stw RC, -4(RA) // Update control var.
4185 | b <3
4186 |
4187 |7: // Skip holes in hash part.
4188 | addi RC, RC, 1
4189 | b <6
4190 break;
4191
4192 case BC_ISNEXT:
4193 | // RA = base*8, RD = target (points to ITERN)
4194 | add RA, BASE, RA
4195 | lwz TMP0, -24(RA)
4196 | lwz CFUNC:TMP1, -20(RA)
4197 | lwz TMP2, -16(RA)
4198 | lwz TMP3, -8(RA)
4199 | cmpwi cr0, TMP2, LJ_TTAB
4200 | cmpwi cr1, TMP0, LJ_TFUNC
4201 | cmpwi cr6, TMP3, LJ_TNIL
4202 | bne cr1, >5
4203 | lbz TMP1, CFUNC:TMP1->ffid
4204 | crand 4*cr0+eq, 4*cr0+eq, 4*cr6+eq
4205 | cmpwi cr7, TMP1, FF_next_N
4206 | srwi TMP0, RD, 1
4207 | crand 4*cr0+eq, 4*cr0+eq, 4*cr7+eq
4208 | add TMP3, PC, TMP0
4209 | bne cr0, >5
4210 | stw ZERO, -4(RA) // Initialize control var.
4211 | addis PC, TMP3, -(BCBIAS_J*4 >> 16)
4212 |1:
4213 | ins_next
4214 |5: // Despecialize bytecode if any of the checks fail.
4215 | li TMP0, BC_JMP
4216 | li TMP1, BC_ITERC
4217 | stb TMP0, -1(PC)
4218 | addis PC, TMP3, -(BCBIAS_J*4 >> 16)
4219 | stb TMP1, 3(PC)
4220 | b <1
4221 break;
4222
4223 case BC_VARG:
4224 | // RA = base*8, RB = (nresults+1)*8, RC = numparams*8
4225 | lwz TMP0, FRAME_PC(BASE)
4226 | add RC, BASE, RC
4227 | add RA, BASE, RA
4228 | addi RC, RC, FRAME_VARG
4229 | add TMP2, RA, RB
4230 | subi TMP3, BASE, 8 // TMP3 = vtop
4231 | sub RC, RC, TMP0 // RC = vbase
4232 | // Note: RC may now be even _above_ BASE if nargs was < numparams.
4233 | cmplwi cr1, RB, 0
4234 | sub. TMP1, TMP3, RC
4235 | beq cr1, >5 // Copy all varargs?
4236 | subi TMP2, TMP2, 16
4237 | ble >2 // No vararg slots?
4238 |1: // Copy vararg slots to destination slots.
4239 | lfd f0, 0(RC)
4240 | addi RC, RC, 8
4241 | stfd f0, 0(RA)
4242 | cmplw RA, TMP2
4243 | cmplw cr1, RC, TMP3
4244 | bge >3 // All destination slots filled?
4245 | addi RA, RA, 8
4246 | blt cr1, <1 // More vararg slots?
4247 |2: // Fill up remainder with nil.
4248 | stw TISNIL, 0(RA)
4249 | cmplw RA, TMP2
4250 | addi RA, RA, 8
4251 | blt <2
4252 |3:
4253 | ins_next
4254 |
4255 |5: // Copy all varargs.
4256 | lwz TMP0, L->maxstack
4257 | li MULTRES, 8 // MULTRES = (0+1)*8
4258 | bley <3 // No vararg slots?
4259 | add TMP2, RA, TMP1
4260 | cmplw TMP2, TMP0
4261 | addi MULTRES, TMP1, 8
4262 | bgt >7
4263 |6:
4264 | lfd f0, 0(RC)
4265 | addi RC, RC, 8
4266 | stfd f0, 0(RA)
4267 | cmplw RC, TMP3
4268 | addi RA, RA, 8
4269 | blt <6 // More vararg slots?
4270 | b <3
4271 |
4272 |7: // Grow stack for varargs.
4273 | mr CARG1, L
4274 | stw RA, L->top
4275 | sub SAVE0, RC, BASE // Need delta, because BASE may change.
4276 | stw BASE, L->base
4277 | sub RA, RA, BASE
4278 | stw PC, SAVE_PC
4279 | srwi CARG2, TMP1, 3
4280 | bl extern lj_state_growstack // (lua_State *L, int n)
4281 | lwz BASE, L->base
4282 | add RA, BASE, RA
4283 | add RC, BASE, SAVE0
4284 | subi TMP3, BASE, 8
4285 | b <6
4286 break;
4287
4288 /* -- Returns ----------------------------------------------------------- */
4289
4290 case BC_RETM:
4291 | // RA = results*8, RD = extra_nresults*8
4292 | add RD, RD, MULTRES // MULTRES >= 8, so RD >= 8.
4293 | // Fall through. Assumes BC_RET follows.
4294 break;
4295
4296 case BC_RET:
4297 | // RA = results*8, RD = (nresults+1)*8
4298 | lwz PC, FRAME_PC(BASE)
4299 | add RA, BASE, RA
4300 | mr MULTRES, RD
4301 |1:
4302 | andi. TMP0, PC, FRAME_TYPE
4303 | xori TMP1, PC, FRAME_VARG
4304 | bne ->BC_RETV_Z
4305 |
4306 |->BC_RET_Z:
4307 | // BASE = base, RA = resultptr, RD = (nresults+1)*8, PC = return
4308 | lwz INS, -4(PC)
4309 | cmpwi RD, 8
4310 | subi TMP2, BASE, 8
4311 | subi RC, RD, 8
4312 | decode_RB8 RB, INS
4313 | beq >3
4314 | li TMP1, 0
4315 |2:
4316 | addi TMP3, TMP1, 8
4317 | lfdx f0, RA, TMP1
4318 | cmpw TMP3, RC
4319 | stfdx f0, TMP2, TMP1
4320 | beq >3
4321 | addi TMP1, TMP3, 8
4322 | lfdx f1, RA, TMP3
4323 | cmpw TMP1, RC
4324 | stfdx f1, TMP2, TMP3
4325 | bne <2
4326 |3:
4327 |5:
4328 | cmplw RB, RD
4329 | decode_RA8 RA, INS
4330 | bgt >6
4331 | sub BASE, TMP2, RA
4332 | lwz LFUNC:TMP1, FRAME_FUNC(BASE)
4333 | ins_next1
4334 | lwz TMP1, LFUNC:TMP1->pc
4335 | lwz KBASE, PC2PROTO(k)(TMP1)
4336 | ins_next2
4337 |
4338 |6: // Fill up results with nil.
4339 | subi TMP1, RD, 8
4340 | addi RD, RD, 8
4341 | stwx TISNIL, TMP2, TMP1
4342 | b <5
4343 |
4344 |->BC_RETV_Z: // Non-standard return case.
4345 | andi. TMP2, TMP1, FRAME_TYPEP
4346 | bne ->vm_return
4347 | // Return from vararg function: relocate BASE down.
4348 | sub BASE, BASE, TMP1
4349 | lwz PC, FRAME_PC(BASE)
4350 | b <1
4351 break;
4352
4353 case BC_RET0: case BC_RET1:
4354 | // RA = results*8, RD = (nresults+1)*8
4355 | lwz PC, FRAME_PC(BASE)
4356 | add RA, BASE, RA
4357 | mr MULTRES, RD
4358 | andi. TMP0, PC, FRAME_TYPE
4359 | xori TMP1, PC, FRAME_VARG
4360 | bney ->BC_RETV_Z
4361 |
4362 | lwz INS, -4(PC)
4363 | subi TMP2, BASE, 8
4364 | decode_RB8 RB, INS
4365 if (op == BC_RET1) {
4366 | lfd f0, 0(RA)
4367 | stfd f0, 0(TMP2)
4368 }
4369 |5:
4370 | cmplw RB, RD
4371 | decode_RA8 RA, INS
4372 | bgt >6
4373 | sub BASE, TMP2, RA
4374 | lwz LFUNC:TMP1, FRAME_FUNC(BASE)
4375 | ins_next1
4376 | lwz TMP1, LFUNC:TMP1->pc
4377 | lwz KBASE, PC2PROTO(k)(TMP1)
4378 | ins_next2
4379 |
4380 |6: // Fill up results with nil.
4381 | subi TMP1, RD, 8
4382 | addi RD, RD, 8
4383 | stwx TISNIL, TMP2, TMP1
4384 | b <5
4385 break;
4386
4387 /* -- Loops and branches ------------------------------------------------ */
4388
4389 case BC_FORL:
4390#if LJ_HASJIT
4391 | hotloop
4392#endif
4393 | // Fall through. Assumes BC_IFORL follows.
4394 break;
4395
4396 case BC_JFORI:
4397 case BC_JFORL:
4398#if !LJ_HASJIT
4399 break;
4400#endif
4401 case BC_FORI:
4402 case BC_IFORL:
4403 | // RA = base*8, RD = target (after end of loop or start of loop)
4404 vk = (op == BC_IFORL || op == BC_JFORL);
4405 if (LJ_DUALNUM) {
4406 | // Integer loop.
4407 | lwzux TMP1, RA, BASE
4408 | lwz CARG1, FORL_IDX*8+4(RA)
4409 | cmplw cr0, TMP1, TISNUM
4410 if (vk) {
4411 | lwz CARG3, FORL_STEP*8+4(RA)
4412 | bne >9
4413 | addo. CARG1, CARG1, CARG3
4414 | cmpwi cr6, CARG3, 0
4415 | lwz CARG2, FORL_STOP*8+4(RA)
4416 | bso >6
4417 |4:
4418 | stw CARG1, FORL_IDX*8+4(RA)
4419 } else {
4420 | lwz TMP3, FORL_STEP*8(RA)
4421 | lwz CARG3, FORL_STEP*8+4(RA)
4422 | lwz TMP2, FORL_STOP*8(RA)
4423 | lwz CARG2, FORL_STOP*8+4(RA)
4424 | cmplw cr7, TMP3, TISNUM
4425 | cmplw cr1, TMP2, TISNUM
4426 | crand 4*cr0+eq, 4*cr0+eq, 4*cr7+eq
4427 | crand 4*cr0+eq, 4*cr0+eq, 4*cr1+eq
4428 | cmpwi cr6, CARG3, 0
4429 | bne >9
4430 }
4431 | blt cr6, >5
4432 | cmpw CARG1, CARG2
4433 |1:
4434 | stw TISNUM, FORL_EXT*8(RA)
4435 if (op != BC_JFORL) {
4436 | srwi RD, RD, 1
4437 }
4438 | stw CARG1, FORL_EXT*8+4(RA)
4439 if (op != BC_JFORL) {
4440 | add RD, PC, RD
4441 }
4442 if (op == BC_FORI) {
4443 | bgt >3 // See FP loop below.
4444 } else if (op == BC_JFORI) {
4445 | addis PC, RD, -(BCBIAS_J*4 >> 16)
4446 | bley >7
4447 } else if (op == BC_IFORL) {
4448 | bgt >2
4449 | addis PC, RD, -(BCBIAS_J*4 >> 16)
4450 } else {
4451 | bley =>BC_JLOOP
4452 }
4453 |2:
4454 | ins_next
4455 |5: // Invert check for negative step.
4456 | cmpw CARG2, CARG1
4457 | b <1
4458 if (vk) {
4459 |6: // Potential overflow.
4460 | mcrxr cr0; bley <4 // Ignore unrelated overflow.
4461 | b <2
4462 }
4463 }
4464 if (vk) {
4465 if (LJ_DUALNUM) {
4466 |9: // FP loop.
4467 | lfd f1, FORL_IDX*8(RA)
4468 } else {
4469 | lfdux f1, RA, BASE
4470 }
4471 | lfd f3, FORL_STEP*8(RA)
4472 | lfd f2, FORL_STOP*8(RA)
4473 | lwz TMP3, FORL_STEP*8(RA)
4474 | fadd f1, f1, f3
4475 | stfd f1, FORL_IDX*8(RA)
4476 } else {
4477 if (LJ_DUALNUM) {
4478 |9: // FP loop.
4479 } else {
4480 | lwzux TMP1, RA, BASE
4481 | lwz TMP3, FORL_STEP*8(RA)
4482 | lwz TMP2, FORL_STOP*8(RA)
4483 | cmplw cr0, TMP1, TISNUM
4484 | cmplw cr7, TMP3, TISNUM
4485 | cmplw cr1, TMP2, TISNUM
4486 }
4487 | lfd f1, FORL_IDX*8(RA)
4488 | crand 4*cr0+lt, 4*cr0+lt, 4*cr7+lt
4489 | crand 4*cr0+lt, 4*cr0+lt, 4*cr1+lt
4490 | lfd f2, FORL_STOP*8(RA)
4491 | bge ->vmeta_for
4492 }
4493 | cmpwi cr6, TMP3, 0
4494 if (op != BC_JFORL) {
4495 | srwi RD, RD, 1
4496 }
4497 | stfd f1, FORL_EXT*8(RA)
4498 if (op != BC_JFORL) {
4499 | add RD, PC, RD
4500 }
4501 | fcmpu cr0, f1, f2
4502 if (op == BC_JFORI) {
4503 | addis PC, RD, -(BCBIAS_J*4 >> 16)
4504 }
4505 | blt cr6, >5
4506 if (op == BC_FORI) {
4507 | bgt >3
4508 } else if (op == BC_IFORL) {
4509 if (LJ_DUALNUM) {
4510 | bgty <2
4511 } else {
4512 | bgt >2
4513 }
4514 |1:
4515 | addis PC, RD, -(BCBIAS_J*4 >> 16)
4516 } else if (op == BC_JFORI) {
4517 | bley >7
4518 } else {
4519 | bley =>BC_JLOOP
4520 }
4521 if (LJ_DUALNUM) {
4522 | b <2
4523 } else {
4524 |2:
4525 | ins_next
4526 }
4527 |5: // Negative step.
4528 if (op == BC_FORI) {
4529 | bge <2
4530 |3: // Used by integer loop, too.
4531 | addis PC, RD, -(BCBIAS_J*4 >> 16)
4532 } else if (op == BC_IFORL) {
4533 | bgey <1
4534 } else if (op == BC_JFORI) {
4535 | bgey >7
4536 } else {
4537 | bgey =>BC_JLOOP
4538 }
4539 | b <2
4540 if (op == BC_JFORI) {
4541 |7:
4542 | lwz INS, -4(PC)
4543 | decode_RD8 RD, INS
4544 | b =>BC_JLOOP
4545 }
4546 break;
4547
4548 case BC_ITERL:
4549#if LJ_HASJIT
4550 | hotloop
4551#endif
4552 | // Fall through. Assumes BC_IITERL follows.
4553 break;
4554
4555 case BC_JITERL:
4556#if !LJ_HASJIT
4557 break;
4558#endif
4559 case BC_IITERL:
4560 | // RA = base*8, RD = target
4561 | lwzux TMP1, RA, BASE
4562 | lwz TMP2, 4(RA)
4563 | checknil TMP1; beq >1 // Stop if iterator returned nil.
4564 if (op == BC_JITERL) {
4565 | stw TMP1, -8(RA)
4566 | stw TMP2, -4(RA)
4567 | b =>BC_JLOOP
4568 } else {
4569 | branch_RD // Otherwise save control var + branch.
4570 | stw TMP1, -8(RA)
4571 | stw TMP2, -4(RA)
4572 }
4573 |1:
4574 | ins_next
4575 break;
4576
4577 case BC_LOOP:
4578 | // RA = base*8, RD = target (loop extent)
4579 | // Note: RA/RD is only used by trace recorder to determine scope/extent
4580 | // This opcode does NOT jump, it's only purpose is to detect a hot loop.
4581#if LJ_HASJIT
4582 | hotloop
4583#endif
4584 | // Fall through. Assumes BC_ILOOP follows.
4585 break;
4586
4587 case BC_ILOOP:
4588 | // RA = base*8, RD = target (loop extent)
4589 | ins_next
4590 break;
4591
4592 case BC_JLOOP:
4593#if LJ_HASJIT
4594 | // RA = base*8 (ignored), RD = traceno*8
4595 | lwz TMP1, DISPATCH_J(trace)(DISPATCH)
4596 | srwi RD, RD, 1
4597 | // Traces on PPC don't store the trace number, so use 0.
4598 | stw ZERO, DISPATCH_GL(vmstate)(DISPATCH)
4599 | lwzx TRACE:TMP2, TMP1, RD
4600 | mcrxr cr0 // Clear SO flag.
4601 | lwz TMP2, TRACE:TMP2->mcode
4602 | stw BASE, DISPATCH_GL(jit_base)(DISPATCH)
4603 | mtctr TMP2
4604 | stw L, DISPATCH_GL(jit_L)(DISPATCH)
4605 | addi JGL, DISPATCH, GG_DISP2G+32768
4606 | bctr
4607#endif
4608 break;
4609
4610 case BC_JMP:
4611 | // RA = base*8 (only used by trace recorder), RD = target
4612 | branch_RD
4613 | ins_next
4614 break;
4615
4616 /* -- Function headers -------------------------------------------------- */
4617
4618 case BC_FUNCF:
4619#if LJ_HASJIT
4620 | hotcall
4621#endif
4622 case BC_FUNCV: /* NYI: compiled vararg functions. */
4623 | // Fall through. Assumes BC_IFUNCF/BC_IFUNCV follow.
4624 break;
4625
4626 case BC_JFUNCF:
4627#if !LJ_HASJIT
4628 break;
4629#endif
4630 case BC_IFUNCF:
4631 | // BASE = new base, RA = BASE+framesize*8, RB = LFUNC, RC = nargs*8
4632 | lwz TMP2, L->maxstack
4633 | lbz TMP1, -4+PC2PROTO(numparams)(PC)
4634 | lwz KBASE, -4+PC2PROTO(k)(PC)
4635 | cmplw RA, TMP2
4636 | slwi TMP1, TMP1, 3
4637 | bgt ->vm_growstack_l
4638 if (op != BC_JFUNCF) {
4639 | ins_next1
4640 }
4641 |2:
4642 | cmplw NARGS8:RC, TMP1 // Check for missing parameters.
4643 | blt >3
4644 if (op == BC_JFUNCF) {
4645 | decode_RD8 RD, INS
4646 | b =>BC_JLOOP
4647 } else {
4648 | ins_next2
4649 }
4650 |
4651 |3: // Clear missing parameters.
4652 | stwx TISNIL, BASE, NARGS8:RC
4653 | addi NARGS8:RC, NARGS8:RC, 8
4654 | b <2
4655 break;
4656
4657 case BC_JFUNCV:
4658#if !LJ_HASJIT
4659 break;
4660#endif
4661 | NYI // NYI: compiled vararg functions
4662 break; /* NYI: compiled vararg functions. */
4663
4664 case BC_IFUNCV:
4665 | // BASE = new base, RA = BASE+framesize*8, RB = LFUNC, RC = nargs*8
4666 | lwz TMP2, L->maxstack
4667 | add TMP1, BASE, RC
4668 | add TMP0, RA, RC
4669 | stw LFUNC:RB, 4(TMP1) // Store copy of LFUNC.
4670 | addi TMP3, RC, 8+FRAME_VARG
4671 | lwz KBASE, -4+PC2PROTO(k)(PC)
4672 | cmplw TMP0, TMP2
4673 | stw TMP3, 0(TMP1) // Store delta + FRAME_VARG.
4674 | bge ->vm_growstack_l
4675 | lbz TMP2, -4+PC2PROTO(numparams)(PC)
4676 | mr RA, BASE
4677 | mr RC, TMP1
4678 | ins_next1
4679 | cmpwi TMP2, 0
4680 | addi BASE, TMP1, 8
4681 | beq >3
4682 |1:
4683 | cmplw RA, RC // Less args than parameters?
4684 | lwz TMP0, 0(RA)
4685 | lwz TMP3, 4(RA)
4686 | bge >4
4687 | stw TISNIL, 0(RA) // Clear old fixarg slot (help the GC).
4688 | addi RA, RA, 8
4689 |2:
4690 | addic. TMP2, TMP2, -1
4691 | stw TMP0, 8(TMP1)
4692 | stw TMP3, 12(TMP1)
4693 | addi TMP1, TMP1, 8
4694 | bne <1
4695 |3:
4696 | ins_next2
4697 |
4698 |4: // Clear missing parameters.
4699 | li TMP0, LJ_TNIL
4700 | b <2
4701 break;
4702
4703 case BC_FUNCC:
4704 case BC_FUNCCW:
4705 | // BASE = new base, RA = BASE+framesize*8, RB = CFUNC, RC = nargs*8
4706 if (op == BC_FUNCC) {
4707 | lwz TMP3, CFUNC:RB->f
4708 } else {
4709 | lwz TMP3, DISPATCH_GL(wrapf)(DISPATCH)
4710 }
4711 | add TMP1, RA, NARGS8:RC
4712 | lwz TMP2, L->maxstack
4713 | add RC, BASE, NARGS8:RC
4714 | stw BASE, L->base
4715 | cmplw TMP1, TMP2
4716 | stw RC, L->top
4717 | li_vmstate C
4718 | mtctr TMP3
4719 if (op == BC_FUNCCW) {
4720 | lwz CARG2, CFUNC:RB->f
4721 }
4722 | mr CARG1, L
4723 | bgt ->vm_growstack_c // Need to grow stack.
4724 | st_vmstate
4725 | bctrl // (lua_State *L [, lua_CFunction f])
4726 | // Returns nresults.
4727 | lwz BASE, L->base
4728 | slwi RD, CRET1, 3
4729 | lwz TMP1, L->top
4730 | li_vmstate INTERP
4731 | lwz PC, FRAME_PC(BASE) // Fetch PC of caller.
4732 | sub RA, TMP1, RD // RA = L->top - nresults*8
4733 | st_vmstate
4734 | b ->vm_returnc
4735 break;
4736
4737 /* ---------------------------------------------------------------------- */
4738
4739 default:
4740 fprintf(stderr, "Error: undefined opcode BC_%s\n", bc_names[op]);
4741 exit(2);
4742 break;
4743 }
4744}
4745
4746static int build_backend(BuildCtx *ctx)
4747{
4748 int op;
4749
4750 dasm_growpc(Dst, BC__MAX);
4751
4752 build_subroutines(ctx);
4753
4754 |.code_op
4755 for (op = 0; op < BC__MAX; op++)
4756 build_ins(ctx, (BCOp)op, op);
4757
4758 return BC__MAX;
4759}
4760
4761/* Emit pseudo frame-info for all assembler functions. */
4762static void emit_asm_debug(BuildCtx *ctx)
4763{
4764 int fcofs = (int)((uint8_t *)ctx->glob[GLOB_vm_ffi_call] - ctx->code);
4765 int i;
4766 switch (ctx->mode) {
4767 case BUILD_elfasm:
4768 fprintf(ctx->fp, "\t.section .debug_frame,\"\",@progbits\n");
4769 fprintf(ctx->fp,
4770 ".Lframe0:\n"
4771 "\t.long .LECIE0-.LSCIE0\n"
4772 ".LSCIE0:\n"
4773 "\t.long 0xffffffff\n"
4774 "\t.byte 0x1\n"
4775 "\t.string \"\"\n"
4776 "\t.uleb128 0x1\n"
4777 "\t.sleb128 -4\n"
4778 "\t.byte 65\n"
4779 "\t.byte 0xc\n\t.uleb128 1\n\t.uleb128 0\n"
4780 "\t.align 2\n"
4781 ".LECIE0:\n\n");
4782 fprintf(ctx->fp,
4783 ".LSFDE0:\n"
4784 "\t.long .LEFDE0-.LASFDE0\n"
4785 ".LASFDE0:\n"
4786 "\t.long .Lframe0\n"
4787 "\t.long .Lbegin\n"
4788 "\t.long %d\n"
4789 "\t.byte 0xe\n\t.uleb128 %d\n"
4790 "\t.byte 0x11\n\t.uleb128 65\n\t.sleb128 -1\n"
4791 "\t.byte 0x5\n\t.uleb128 70\n\t.uleb128 55\n",
4792 fcofs, CFRAME_SIZE);
4793 for (i = 14; i <= 31; i++)
4794 fprintf(ctx->fp,
4795 "\t.byte %d\n\t.uleb128 %d\n"
4796 "\t.byte %d\n\t.uleb128 %d\n",
4797 0x80+i, 37+(31-i), 0x80+32+i, 2+2*(31-i));
4798 fprintf(ctx->fp,
4799 "\t.align 2\n"
4800 ".LEFDE0:\n\n");
4801#if LJ_HASFFI
4802 fprintf(ctx->fp,
4803 ".LSFDE1:\n"
4804 "\t.long .LEFDE1-.LASFDE1\n"
4805 ".LASFDE1:\n"
4806 "\t.long .Lframe0\n"
4807 "\t.long lj_vm_ffi_call\n"
4808 "\t.long %d\n"
4809 "\t.byte 0x11\n\t.uleb128 65\n\t.sleb128 -1\n"
4810 "\t.byte 0x8e\n\t.uleb128 2\n"
4811 "\t.byte 0xd\n\t.uleb128 0xe\n"
4812 "\t.align 2\n"
4813 ".LEFDE1:\n\n", (int)ctx->codesz - fcofs);
4814#endif
4815 fprintf(ctx->fp, "\t.section .eh_frame,\"a\",@progbits\n");
4816 fprintf(ctx->fp,
4817 ".Lframe1:\n"
4818 "\t.long .LECIE1-.LSCIE1\n"
4819 ".LSCIE1:\n"
4820 "\t.long 0\n"
4821 "\t.byte 0x1\n"
4822 "\t.string \"zPR\"\n"
4823 "\t.uleb128 0x1\n"
4824 "\t.sleb128 -4\n"
4825 "\t.byte 65\n"
4826 "\t.uleb128 6\n" /* augmentation length */
4827 "\t.byte 0x1b\n" /* pcrel|sdata4 */
4828 "\t.long lj_err_unwind_dwarf-.\n"
4829 "\t.byte 0x1b\n" /* pcrel|sdata4 */
4830 "\t.byte 0xc\n\t.uleb128 1\n\t.uleb128 0\n"
4831 "\t.align 2\n"
4832 ".LECIE1:\n\n");
4833 fprintf(ctx->fp,
4834 ".LSFDE2:\n"
4835 "\t.long .LEFDE2-.LASFDE2\n"
4836 ".LASFDE2:\n"
4837 "\t.long .LASFDE2-.Lframe1\n"
4838 "\t.long .Lbegin-.\n"
4839 "\t.long %d\n"
4840 "\t.uleb128 0\n" /* augmentation length */
4841 "\t.byte 0xe\n\t.uleb128 %d\n"
4842 "\t.byte 0x11\n\t.uleb128 65\n\t.sleb128 -1\n"
4843 "\t.byte 0x5\n\t.uleb128 70\n\t.uleb128 55\n",
4844 fcofs, CFRAME_SIZE);
4845 for (i = 14; i <= 31; i++)
4846 fprintf(ctx->fp,
4847 "\t.byte %d\n\t.uleb128 %d\n"
4848 "\t.byte %d\n\t.uleb128 %d\n",
4849 0x80+i, 37+(31-i), 0x80+32+i, 2+2*(31-i));
4850 fprintf(ctx->fp,
4851 "\t.align 2\n"
4852 ".LEFDE2:\n\n");
4853#if LJ_HASFFI
4854 fprintf(ctx->fp,
4855 ".Lframe2:\n"
4856 "\t.long .LECIE2-.LSCIE2\n"
4857 ".LSCIE2:\n"
4858 "\t.long 0\n"
4859 "\t.byte 0x1\n"
4860 "\t.string \"zR\"\n"
4861 "\t.uleb128 0x1\n"
4862 "\t.sleb128 -4\n"
4863 "\t.byte 65\n"
4864 "\t.uleb128 1\n" /* augmentation length */
4865 "\t.byte 0x1b\n" /* pcrel|sdata4 */
4866 "\t.byte 0xc\n\t.uleb128 1\n\t.uleb128 0\n"
4867 "\t.align 2\n"
4868 ".LECIE2:\n\n");
4869 fprintf(ctx->fp,
4870 ".LSFDE3:\n"
4871 "\t.long .LEFDE3-.LASFDE3\n"
4872 ".LASFDE3:\n"
4873 "\t.long .LASFDE3-.Lframe2\n"
4874 "\t.long lj_vm_ffi_call-.\n"
4875 "\t.long %d\n"
4876 "\t.uleb128 0\n" /* augmentation length */
4877 "\t.byte 0x11\n\t.uleb128 65\n\t.sleb128 -1\n"
4878 "\t.byte 0x8e\n\t.uleb128 2\n"
4879 "\t.byte 0xd\n\t.uleb128 0xe\n"
4880 "\t.align 2\n"
4881 ".LEFDE3:\n\n", (int)ctx->codesz - fcofs);
4882#endif
4883 break;
4884 default:
4885 break;
4886 }
4887}
4888
generated by cgit v1.2.3-55-g6feb (git 2.39.1) at 2025-08-11 03:55:16 +0000